A dynamic target optimization control method for the decocting process of a veterinary traditional Chinese medicine compound

Abstract

The application provides a dynamic target optimization control method for a veterinary traditional Chinese medicine compound decoction process, and relates to the technical field of process control.The method comprises the following steps: constructing a fluid dynamics characteristic space, using a detrended and filtered pressure time series signal, calculating a pressure fluctuation variance index to represent the turbulence intensity; constructing a thermodynamic energy observation space, using an extended Kalman filter to solve the system thermal efficiency coefficient in real time; performing time domain lag compensation, calculating the partial derivative eigenvalue of the thermal efficiency relative to the turbulence intensity, and generating a fluid-thermal coupling sensitivity index; inputting the index into a dynamic extreme value search controller based on phase plane trajectory constraints to generate an adaptive power instruction. Through driving the sensitivity index to converge to a zero value dead zone, the application realizes dynamic impedance matching of physical disturbance and chemical endothermic of the medicine liquid, and solves the problems of energy efficiency feedback lag and extraction working condition mismatch in traditional control.

Description

Technical Field

[0001] This invention relates to the field of process control technology, specifically to a dynamic target optimization control method for the decoction process of veterinary traditional Chinese medicine compound. Background Technology

[0002] The decoction process of traditional Chinese medicine compound for veterinary use is a typical nonlinear, time-delayed heat and mass transfer process accompanied by complex biochemical reactions. In modern pharmaceutical engineering and automation control, achieving optimal energy efficiency and uniform quality in this process has always been a key technological challenge. Existing control strategies for this process generally suffer from the limitation of separating the "physical state" from the "chemical process." Specifically, mainstream solutions often use pressure or level feedback to maintain a constant boiling state, but "high turbulence" (violent boiling) in the physical sense is not equivalent to "efficient extraction" in the chemical sense. When the liquid reaches dissolution saturation, cavitation heat resistance occurs at the solid-liquid interface, or the hygroscopicity of the medicinal materials changes over time, continuously increasing physical disturbances not only fails to improve the extraction rate of effective components but also leads to a surge in ineffective energy consumption and the destruction of heat-sensitive components. The dissolution and transformation of the drug's effective components are accompanied by specific endothermic or exothermic reactions (enthalpy change), but this process cannot be directly sensed by conventional temperature sensors. Existing constant power or constant temperature control lacks a real-time feedback mechanism for "effective energy absorption efficiency." Although existing technologies (such as CN121143132A) propose a control method based on predictive models and offline benchmarks, which solves the problems of data privacy and basic operating condition control through security boundaries and federated learning, they still rely on the static mapping of offline data in essence. They are difficult to perceive and respond to the rapidly changing fluid-thermal dynamic coupling characteristics during the cooking process in real time, and cannot establish a real-time marginal benefit evaluation mechanism between "physical disturbance input" and "chemical extraction output".

[0003] To address the aforementioned limitations, there is an urgent need in this field for control methods that can break down the boundaries between physical and chemical fields. This method constructs a coupled fusion index of fluid dynamics and thermodynamics, transforming traditional "steady-state control" into "dynamic extremum optimization control." Specifically, by calculating the fluid-thermal coupling sensitivity index in real time, it achieves fluid-thermal dynamic impedance matching during the decoction process, thereby fundamentally solving the technical problems of energy efficiency mismatch and low extraction efficiency. Summary of the Invention

[0004] The purpose of this invention is to provide a dynamic target optimization control method for the decoction process of compound traditional Chinese medicine for veterinary use, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A dynamic target optimization control method for the decoction process of a compound traditional Chinese medicine for veterinary use, comprising the following steps:

[0007] S101: Constructing the fluid dynamics feature space

[0008] Within a continuous time window, the pressure time-series signal inside the decoction container is acquired; the pressure time-series signal is subjected to detrending processing and frequency band filtering, and first state data characterizing the current physical turbulence mixing intensity of the medicinal liquid is calculated and generated.

[0009] S201: Constructing a thermodynamic energy observation space

[0010] The input energy data of the heating actuator acting on the decoction container and the temperature rise response data of the medicinal liquid are acquired; using a preset energy conservation state observer, based on the residual between the input energy data and the temperature rise response data, the second state data characterizing the current instantaneous effective energy absorption efficiency of the medicinal liquid is calculated in real time and generated.

[0011] S301: Generative Fluid-Thermodynamic Coupling Fusion Indicator

[0012] Perform time-domain hysteresis compensation alignment on the first state data and the second state data; calculate the partial derivative eigenvalue of the rate of change of the second state data relative to the rate of change of the first state data at the current moment, and define the partial derivative eigenvalue as the fluid-thermal coupling sensitivity index;

[0013] S401: Perform dynamic extreme value search control

[0014] The fluid-thermal coupling sensitivity index is input as a controlled variable to the dynamic extreme value search controller; the dynamic extreme value search controller generates control commands for adjusting the power of the heating actuator based on gradient descent logic, so as to drive the fluid-thermal coupling sensitivity index to converge to a preset dead zone near zero, thereby locking the optimal impedance matching point of turbulence-endothermic reaction of the liquid under the current operating conditions.

[0015] Compared with the prior art, the beneficial effects of the present invention are:

[0016] This invention quantifies the marginal gain of the "rate of change of thermal efficiency" relative to the "rate of change of turbulence intensity" by constructing a fluid-thermal coupling sensitivity index (step S301) and calculating it using time-domain hysteresis compensation and partial derivative eigenvalues. This technique overcomes the blindness of existing technologies that rely solely on physical quantities (pressure / temperature) for feedback. Its mechanism lies in the fact that this index replicates the law of diminishing marginal utility in economics, accurately identifying three operating conditions: "under-stirring" (positive gain), "optimal matching" (gain approaching zero), and "cavitation resistance" (negative gain). Based on this, the sensitivity index converges to the zero-value dead zone, thereby locking in the optimal turbulence-endothermic equilibrium point of the liquid under the current operating conditions, avoiding energy waste and drug efficacy loss caused by ineffective boiling.

[0017] This invention employs dynamic extremum search control based on phase plane trajectory constraints (step S401), constructing a two-dimensional phase plane state space with the sensitivity exponent and its differential components as dimensions. It also utilizes the generalized tracking error potential energy to generate an adaptive gain modulation factor, solving the control divergence problem under complex nonlinear conditions and enhancing the system's robustness and convergence speed. This technique overcomes the oscillation or slow convergence problems that easily occur in traditional PID control or fixed-step gradient search in nonlinear time-varying systems. The solution mechanism lies in: through the nonlinear mapping of the hyperbolic tangent function, adaptive adjustment of the control step size is achieved, with large step sizes rapidly approaching the extreme point when far away, and decaying exponentially when approaching the zero dead zone. This variable step size mechanism, combined with the robustness degradation strategy, effectively suppresses control divergence caused by sensor noise or sudden changes in operating conditions (false boiling), ensuring stable system operation.

[0018] This invention utilizes an energy conservation state observer based on Extended Kalman Filter (EKF) (step S201) to establish a nonlinear state equation encompassing sensible heat, heat dissipation, and reaction endothermics, and recursively estimates the system's thermal efficiency coefficient in real time. This achieves precise observation of the invisible thermodynamic state, improving the transparency of process control. This technique overcomes the technical challenge of directly measuring the endothermic heat of chemical reactions and the latent heat of phase transitions in existing technologies. The solution mechanism lies in the fact that the EKF algorithm effectively filters out process noise and observation noise, decoupling the implicit "effective energy absorption efficiency" from the complex temperature response by minimizing the residual. Attached Figure Description

[0019] Figure 1 This is a schematic diagram illustrating the technical principle and logical architecture of a dynamic target optimization control method for the decoction process of veterinary Chinese herbal medicine provided in an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of the technical route of the present invention;

[0021] Figure 3 This is a verification diagram of the adaptive gain response characteristics of fluid-thermal coupling. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Example 1:

[0025] Please see Figures 1 to 3 The present invention provides a technical solution:

[0026] A dynamic target optimization control method for the decoction process of a compound traditional Chinese medicine for veterinary use includes the following steps:

[0027] S101: Constructing the fluid dynamics feature space

[0028] Within a continuous time window, the pressure time-series signal inside the decoction container is acquired; the pressure time-series signal is subjected to detrending processing and frequency band filtering, and first state data characterizing the current physical turbulence mixing intensity of the medicinal liquid is calculated and generated.

[0029] S201: Constructing a thermodynamic energy observation space

[0030] The input energy data of the heating actuator acting on the decoction container and the temperature rise response data of the medicinal liquid are acquired; using a preset energy conservation state observer, based on the residual between the input energy data and the temperature rise response data, the second state data characterizing the current instantaneous effective energy absorption efficiency of the medicinal liquid is calculated in real time and generated.

[0031] S301: Generative Fluid-Thermodynamic Coupling Fusion Indicator

[0032] Perform time-domain hysteresis compensation alignment on the first state data and the second state data; calculate the partial derivative eigenvalue of the rate of change of the second state data relative to the rate of change of the first state data at the current moment, and define the partial derivative eigenvalue as the fluid-thermal coupling sensitivity index;

[0033] S401: Perform dynamic extreme value search control

[0034] The fluid-thermal coupling sensitivity index is input as a controlled variable to the dynamic extreme value search controller; the dynamic extreme value search controller generates control commands for adjusting the power of the heating actuator based on gradient descent logic, so as to drive the fluid-thermal coupling sensitivity index to converge to a preset dead zone near zero, thereby locking the optimal impedance matching point of turbulence-endothermic reaction of the liquid under the current operating conditions.

[0035] This embodiment operates in a control environment with basic data acquisition and processing capabilities. This environment includes, but is not limited to: a pressure sensing module with a sampling rate of no less than 50Hz, a heating execution module with PID or PWM adjustment functions, and a computing core capable of performing floating-point operations.

[0036] Further defining step S101, in step S101, the first state data is the pressure fluctuation variance index, and the calculation and generation process includes:

[0037] The pressure timing signal is extracted using a sliding window algorithm;

[0038] The DC liquid level component and high-frequency electromagnetic noise component in the pressure time series signal are removed by using a bandpass filter, while retaining the bubble dynamics frequency band component that characterizes the bubble formation and rupture features.

[0039] Calculate the statistical variance of the bubble dynamics frequency band components and quantify the statistical variance into the pressure fluctuation variance index.

[0040] For step S101, the specific implementation steps for calculating and generating the pressure fluctuation variance index are as follows:

[0041] In this embodiment, the first state data is defined as the "pressure fluctuation variance index." This index aims to separate macroscopic liquid level changes from high-frequency electromagnetic interference and quantify the sound pressure characteristics caused by the nucleation and collapse of boiling bubbles in the liquid. The specific calculation and generation process is as follows:

[0042] The raw pressure timing signal from the pressure sensor is read via an analog-to-digital converter (ADC) at a frequency that satisfies the Nyquist sampling theorem (preferably not less than 50Hz in this embodiment). A sliding window algorithm is used to extract data, with the sliding window length (5 seconds) set as the observation period to ensure sufficient bubble dynamics characteristic data points are contained within the window, while also guaranteeing the dynamic response speed of this scheme to state changes. The sliding window algorithm employs a first-in-first-out (FIFO) queue mechanism, with a window update step size preferably between 0.1 and 1.0 seconds. This means that the above calculation process is executed every 0.1-1.0 seconds, thereby achieving near real-time monitoring of the boiling state, ensuring both sufficient statistical samples and timely control feedback.

[0043] Regarding the logic for determining the length of the sliding window: Specific implementation method: Construction and solution of the fluid dynamics feature space in step S101;

[0044] In this embodiment, the core objective of step S101 is to generate a "pressure fluctuation variance index" that accurately characterizes the intensity of physical turbulent mixing of the drug solution. To ensure that this index is unaffected by liquid level changes, water addition operations, and electromagnetic noise, the following rigorous data processing procedure is performed, including dynamic window construction, detrending processing, and frequency domain feature extraction:

[0045] A circular data buffer with First-In-First-Out (FIFO) characteristics is established to capture pressure timing signals. To balance the completeness of fluid feature extraction with the system's real-time response capability, the "window length" of this buffer is not fixed but dynamically determined based on fluid dynamics principles: the "reference window length parameter" is read by accessing an external configuration file stored in the non-volatile memory of the control host through a file parsing interface. In the preferred configuration of this embodiment, this parameter is set to 5000 milliseconds.

[0046] The determination of this value follows the following physical calculation logic: Identify the low cutoff frequency parameter set by the bandpass filter (0.5 Hz in this embodiment); calculate the maximum physical fluctuation period corresponding to this frequency (i.e., 2000 milliseconds); introduce a dimensionless "Shannon-Nyquist coverage factor," and multiply the maximum physical fluctuation period by this coverage factor (preferably 2.5 times) to obtain a baseline value of 5000 milliseconds. This setting ensures that each sliding window fully covers at least 2.5 physical cycles of bubble generation and collapse in the time domain, thereby statistically eliminating variance calculation distortion caused by truncation errors. Regardless of container specifications, as long as the dominant physical frequency of stirring or boiling is not lower than 0.5 Hz, this window length ensures that the data accurately reflects the current turbulence intensity of the fluid, rather than instantaneous noise.

[0047] To address the issue of "removing the DC level component," this embodiment employs a linear regression detrending algorithm to eliminate the sloping trend introduced by dynamic water addition or slow level drift. The specific computational steps include: acquiring the original pressure time-series dataset within the current sliding window, defining it as a two-dimensional coordinate point set, where the time axis is the independent variable and the pressure amplitude is the dependent variable. A linear fit is performed on the coordinate point set using the least squares method. By minimizing the sum of squared residuals between the actual pressure value and the fitted line, the slope and intercept of the best-fit line (trend line) are calculated. This trend line physically represents the current static level height or the slow water addition / evaporation trend (DC component). For each original pressure data point within the window, its corresponding theoretical value on the trend line is subtracted. Through this subtraction operation, baseline deviations caused by different level heights or sensor zero-point drift are eliminated, resulting in a zero-mean dynamic pressure fluctuation signal. This signal retains only the AC component related to fluid disturbance.

[0048] The detrended signal is input to a preset digital bandpass filter. In order to "preserve the bubble dynamics frequency band components that characterize bubble formation and rupture", this embodiment physically defines the frequency band: a Butterworth filter is preferably used to take advantage of its flat amplitude-frequency characteristics within the passband and avoid amplitude distortion of the effective signal.

[0049] Low cutoff frequency (0.5Hz): Used to filter out low-frequency macroscopic gravity wave interference (<0.5Hz) caused by the rotation of the agitator, the addition of water to the solution, or the sloshing of the liquid.

[0050] High cutoff frequency (5.0Hz): Used to filter out high-frequency noise (>5Hz) introduced by electromagnetic pump operation, power grid interference or mechanical vibration.

[0051] According to the principles of two-phase flow dynamics, the pressure pulsations caused by the collapse of boiling bubbles are mainly concentrated in the 0.5Hz to 5.0Hz frequency band. Locking onto this frequency band maximizes the boiling signal-to-noise ratio, ensuring that the control system "only hears the sound of bubbles." The waveform data (pure bubble dynamics component) output by the filter is acquired, and the statistical variance of this waveform data within the current window is calculated. This variance value is quantified as the "pressure fluctuation variance index".

[0052] Further specifying step S201, in step S201, the energy conservation state observer is constructed based on the Extended Kalman Filter (EKF) algorithm; the second state data is the system thermal efficiency coefficient, and the calculation process includes:

[0053] Establish a nonlinear thermodynamic equation of state that includes sensible heat terms, heat dissipation loss terms, and reaction endothermic terms;

[0054] The input energy data is used as the control input, and the temperature rise response data is used as the observation output.

[0055] Using the extended Kalman filter algorithm, the implicit state variables in the nonlinear thermodynamic equation of state are recursively estimated by minimizing the observation residuals, thereby obtaining the thermal efficiency coefficient of the system.

[0056] For the specific implementation of step S201: The following are the steps for calculating the system thermal efficiency coefficient based on the Extended Kalman Filter (EKF):

[0057] This embodiment establishes a nonlinear thermodynamic equation of state that includes sensible heat terms, heat dissipation loss terms, and reaction endothermic terms. The system thermal efficiency coefficient is then... Treating these as augmented state variables that drift slowly with the progress of the chemical reaction, the continuous-time equation of state is constructed as follows: ;in: : Characterizes the sensible heat term, where T is the temperature of the drug solution, C is the specific heat capacity, and m is the mass; Characterizes effective energy input. For input electrical power, The system thermal efficiency coefficient to be estimated; : Characterizes the heat dissipation loss term, For heat dissipation coefficient, The ambient temperature; : Characterizes the endothermic term of the reaction, covering the endothermic reaction of chemical reaction and the latent heat of phase change, and is the reaction heat model value preset according to the process stage.

[0058] For the Extended Kalman Filter (EKF) solution process: When acquiring "input energy data" to drive the energy conservation state observer, the following periodic integration preprocessing logic is executed to address the instantaneous value discrepancy caused by pulse width modulation (PWM): Establish the update step size with the energy conservation state observer (hereinafter referred to as the observer). (In this embodiment, it is a 1-second) synchronized energy integrator. Specifically, in each observer update cycle... Within this range, instantaneous voltage u(t) and instantaneous current i(t) are sampled at a high frequency (10kHz), and the accumulated energy within the interval is calculated in real time. :

[0059] ;use Calculate the average effective power within this period. Here, k is the "index of the discrete time step" or "sampling number"; this average effective power is then assigned as the "input energy data" to the thermodynamic equation of state. The variables ensure that the model input accurately reflects the effective work done by the heating actuator per unit time, eliminating the discretization jitter of power input data caused by actuator switching actions, and ensuring the continuity and authenticity of the energy input terms of the thermodynamic model.

[0060] The temperature rise response data is used as the observation output. The specific recursive operation logic is as follows: Obtain the estimated value of the system thermal efficiency coefficient and the liquid temperature state at the previous moment. Based on the above energy conservation physical model, predict the liquid temperature at the current moment. Collect the actual liquid temperature value at the current moment in real time through a temperature sensor. Subtract the predicted temperature value from the actual collected liquid temperature value to obtain the temperature observation residual. This temperature observation residual reflects the deviation between the model prediction and the actual physical process. Calculate the Kalman gain matrix, which dynamically represents the trust weight of the model prediction value and the sensor observation value. Multiply the temperature observation residual by the Kalman gain to obtain the "system thermal efficiency coefficient". "A reverse correction will be performed. If the actual temperature rise is lower than expected and external heat dissipation factors have been ruled out, the energy conservation state observer will automatically adjust downwards." The value represents the amount of input energy being consumed by internal latent heat or reaction endotherm.

[0061] The drug concentration process is highly nonlinear (specific heat capacity, viscosity, and heat of reaction change dynamically with concentration), and industrial environments are subject to process noise (voltage fluctuations) and observation noise (sensor jitter). Compared to the least squares method, EKF can dynamically adjust the confidence level of the model and the confidence level of the measurement, thereby extracting a smooth and realistic trend of efficiency changes in noisy environments. This step maps the changes in endothermic chemical reactions, latent heat of phase transition, and thermal resistance, which cannot be directly measured, into the observable dimensionless parameter "thermal efficiency," achieving a leap from "physical control" to "chemical control."

[0062] To specifically implement the Extended Kalman Filter (EKF) in a digital control system, this embodiment further discloses the transformation process from a continuous-time model to a discrete-time algorithm:

[0063] Assume the system sampling period is (1 second). Define the state vector for discrete time step k. , For the transpose term; where Let k be the system thermal efficiency coefficient at discrete time step k. Let be the temperature of the liquid solution at discrete time step k. Using the Euler method, the aforementioned continuous differential equation is discretized to obtain the state transition equation:

[0064]

[0065] Where: Assumption It is a random walk process. This is process noise; The input power measured at the previous moment; C,m represents the preset specific heat capacity and drug mass. For the pre-calibrated heat dissipation coefficient, These are the reaction heat model values ​​preset according to the process stage.

[0066] It should be added that the heat dissipation coefficient was calibrated in advance through offline natural cooling experiments. To ensure the accuracy of the state observer's calculations; the calibration process of the heat dissipation coefficient includes two stages: experimental data acquisition and model parameter calculation. The specific steps are as follows: In the calibration stage, inject calibration medium (preferably pure water) of equal mass to the rated process volume into the container. Start the heating actuator to raise the medium temperature to the preset maximum operating temperature ( During the static sampling phase, the power input to the heating actuator is cut off. The feeding is stopped, and the stirring device is kept still to simulate a simple heat conduction and natural convection environment. A temperature sensor is used to continuously collect sequence data of the medium temperature decreasing over time at a preset sampling frequency (1Hz) until the temperature drops to a cutoff threshold close to the ambient temperature. In this embodiment, the cutoff threshold is defined as the ambient temperature. Plus a fixed temperature difference To ensure the signal-to-noise ratio of the sampled data, based on Newton's law of cooling, the following calculations are performed on the temperature-time sequence: For each sampling time t in the sequence, the current medium temperature T(t) and the ambient temperature are calculated. The difference is calculated, and the natural logarithm of this difference is performed to generate a logarithmic temperature difference sequence. Using time t as the independent variable and the logarithmic temperature difference sequence Y(t) as the dependent variable, a least squares (LSM) linear regression was performed. A straight line minimizing the sum of squared residuals was fitted, and the absolute value of the slope of this line was extracted and denoted as the cooling rate constant k1. The total heat capacity parameter was then obtained. (This parameter is obtained by weighted summation of the specific heat capacity, mass, and heat capacity of the container liner of the calibration medium). The calculation logic is as follows: ,in To calibrate the specific heat capacity of the medium, To calibrate the quality of the medium, The predicted equivalent heat capacity of the container's inner liner is calculated using... The heat dissipation coefficient is calculated. The result is then stored in the controller's thermodynamic parameter table and used as a fixed-gain term for calculating real-time heat dissipation losses in the EKF (Extended Kalman Filter) operation.

[0067] In this embodiment, the observed variable is the temperature sensor reading. Its observation equation is a linear equation: ;in To observe noise, corresponding to the sensor's measurement error. In the EKF's "prediction step," the Jacobian matrix of the state transition matrix is ​​calculated. This is used for the transfer of the covariance matrix. Based on the above discretization equation, for the state variables... Taking the partial derivative, we get: .

[0068] Based on the above discretization model, the specific implementation steps are detailed as follows: Set the initial state. and the initial covariance matrix Calculating prior state estimates using discrete equations. ;use Update covariance, where Let the process noise covariance matrix be denoted. Calculate the Kalman gain. Where H = [0,1], and R is the measurement noise variance. Using the residuals... Update the state vector and output the final result. .

[0069] It should be noted that the process noise covariance matrix in this embodiment... Defined as a 2×2 diagonal matrix containing two main diagonal elements: thermal efficiency noise variance ( ) and temperature prediction noise variance ( ).

[0070] The determination logic is as follows: This parameter characterizes the random walk drift velocity of the thermal efficiency coefficient. This embodiment considers that during chemical extraction, the changes in the physical properties (viscosity, specific heat) of the drug solution are a slow, continuous process, rather than a sudden abrupt change. In this embodiment, this value is set to... The technical consideration behind this value is to allow for minute corrections to thermal efficiency on the order of one-thousandth per second, thereby filtering out interference from high-frequency power grid fluctuations on efficiency estimation.

[0071] The determination logic is as follows: This parameter characterizes the uncertainty of the thermodynamic model itself (such as uneven heat distribution caused by uneven stirring). This value is set to... Its magnitude is slightly higher than the measurement noise floor of the temperature sensor, to ensure that the filter maintains a dynamic balance between model prediction and sensor observation.

[0072] Further defining step S301, in step S301, the fluid-thermal coupling sensitivity index characterizes the marginal gain of the drug solution energy absorption efficiency relative to the intensity of physical turbulence; its generation process includes:

[0073] Access the preset hysteresis parameter table and read the thermal hysteresis time constant;

[0074] Using the thermal hysteresis time constant, historical backtracking extraction is performed on the time series of the first state data to obtain the hysteresis first state data that has a causal correspondence with the current moment;

[0075] Calculate the current rate of change of the second state data relative to time, and the historical rate of change of the first state data relative to time, respectively.

[0076] Perform a division operation to calculate the ratio of the current rate of change to the historical rate of change, and perform moving average filtering on the ratio to determine the processed value as the fluid-thermal coupling sensitivity index.

[0077] This embodiment performs the following specific implementation steps for generating fluid-thermal coupled fusion indicators:

[0078] In this embodiment, the core task of step S301 is to construct a fluid-thermal coupling sensitivity index (denoted as ) that can characterize the "marginal gain of drug energy absorption efficiency relative to physical turbulence intensity". This index solves the physical problem of the mismatch between "transient fluid disturbance" and "hysteretic response" on the time axis. Through time-domain hysteresis compensation and differential ratio calculation, it achieves causal alignment and quantitative analysis across physical fields.

[0079] "Preset thermal hysteresis time constant" (denoted as Due to the thermal inertia of the heating plate and the thermal conduction delay of the sensor, instantaneous changes in fluid turbulence (S101 first state data) cannot immediately cause changes in thermal efficiency (S201 second state data). If this lag is not compensated for, the calculated sensitivity index will exhibit a reversed causal relationship. In this embodiment, the following offline step response calibration procedure (SOP) is used to determine this, and the results are stored in the control system's configuration file:

[0080] Access the preset hysteresis parameter table and read the thermal hysteresis time constant. The process of constructing the hysteresis parameter table is as follows:

[0081] Key physical variables affecting thermal inertia are selected as index keys. In this embodiment, "liquid level (or drug mass) in the container" is selected as the variable. A set of discrete calibration nodes (20%, 40%, 60%, 80%, 100% rated capacity) covering the lowest working liquid level to the highest working liquid level are set.

[0082] For each calibration node, the corresponding volume of water was added to a standard boiling vessel, maintaining a stable state of heating from room temperature to 50 degrees Celsius (avoiding boiling disturbances). At time [time missing] The power of the heating actuator is instantaneously adjusted from 0% to 100% (step input). The response curve of the liquid temperature sensor is recorded at a high frequency (e.g., 100Hz). The point where the maximum slope tangent line intersects the temperature response curve with the time axis (the horizontal line where the temperature is maintained at the reference value) is recorded as _____. .Will and The time difference is defined as the thermal hysteresis time constant under this operating condition. Repeat the above steps to obtain the thermal hysteresis time constants corresponding to all calibration nodes. Construct a one-dimensional lookup table using the liquid level height as the index (Key) and the corresponding time constant as the value (Value). In this embodiment, after calibration, when the liquid level is 100%, Seconds; when the liquid level is 50%, Seconds. In actual operation, the system obtains the current thermal hysteresis time constant based on the real-time measured liquid level data through table lookup or linear interpolation.

[0083] During real-time operation, a circular buffer or delay queue of length N is established. The first state data (pressure fluctuation variance exponent, denoted as t) at the current time t is received in real-time. ) and second-state data (system thermal efficiency coefficient, denoted as ).

[0084] Alignment logic: Given that fluid disturbance is the cause and thermal efficiency change is the result, the algorithm retrieves from the historical cache. The first state data corresponding to time step This is compared with the second state data corresponding to the current time t. Perform pairing.

[0085] This step ensures that it occurs between the "moment of disturbance generation" and the "moment of thermal effect generated by the disturbance," achieving strict alignment in causal logic. It should be further noted that, to calculate the fluid-thermal coupling sensitivity index, two physical increment variables that are strictly aligned in temporal causality are constructed. This embodiment uses a history-backtracking differential logic based on a circular buffer to obtain the "turbulence increment." "and thermal efficiency increment" The specific calculation process is as follows:

[0086] First, define a fixed calculation step size. (In this embodiment, 1.0 second is preferred), and two first-in-first-out (FIFO) circular data queues are created in memory to store past data. First state data within the time period ( Pressure fluctuation variance index) and second state data ( (System thermal efficiency coefficient).

[0087] Implement thermal efficiency increment Real-time differential calculation:

[0088] For the second state data, perform the differential calculation for the "current moment": read the system thermal efficiency coefficient at the current sampling time t in real time from the output port of the state observer, and denot it as... Read the previous calculation step time from the circular data queue. The system thermal efficiency coefficient is denoted as . ;

[0089] The system thermal efficiency coefficient at the current moment Subtract the system thermal efficiency coefficient at the previous moment The calculated difference is defined as the thermal efficiency increment. This value quantifies the magnitude and direction of energy efficiency changes within the most recent calculation period.

[0090] Lag turbulence increment Historical backtracking differential calculation: For the first state data, perform backtracking differential calculation for the "historical moment": read the preset thermal hysteresis time constant. (15 seconds). By subtracting the thermal hysteresis time constant from the current time t, the historical time can be located. .according to The timestamp is indexed and retrieved in the circular data queue of the first state data to extract the pressure fluctuation variance index corresponding to that historical moment, denoted as . Within the same queue, further backtracking is performed to extract the previous calculation step time of that historical moment (…). The data is denoted as Data from historical moments Subtracting data from previous historical moments The calculated difference is defined as the hysteresis turbulence increment. . It represents "the turbulent changes that occurred at that instant 15 seconds ago," while It represents "the change in thermal efficiency that is only manifested at the current moment as a result." This time-axis aligned differential calculation mathematically eliminates spurious correlations caused by time misalignment, ensuring that the sensitivity index of subsequent calculations truly reflects the physical causal relationship.

[0091] Based on the aligned data, the marginal rate of change of the second-state data relative to the first-state data is calculated. In this embodiment, to avoid the numerical singularity problem when the denominator is zero, a regularized difference ratio algorithm is used. The specific calculation logic is as follows: Calculate the difference between the numerator and the denominator:

[0092] ; ;

[0093] Introducing a preset minimum perturbation resolution threshold r1 (the value in this embodiment) Construct anti-saturation division operation logic:

[0094] ;

[0095] Applying a smoothing filter to the above ratio yields the final fluid-thermal coupling sensitivity index: ;in: To calculate the step size; This is a moving average filter used to remove high-frequency quantization noise introduced by differentiation operations. It is explicitly stated that when the physical turbulence change is not significant (the denominator is below the minimum disturbance resolution threshold r1), the sensitivity is forcibly assumed to be zero, thus ensuring the numerical safety of the control system in steady state.

[0096] Wherein: the numerator term corresponds to the "current rate of change" of the instantaneous increment of thermal efficiency (represented as a difference in the discrete domain), and the denominator term characterizes the instantaneous increment of turbulence intensity after hysteresis compensation, which in turn corresponds to the "historical rate of change".

[0097] When the fluid-thermal coupling sensitivity index (Positive benefit): This indicates that increasing the heat output (increasing turbulence) can effectively improve thermal efficiency. The system is in a state of "insufficient stirring", and the control strategy should continue to heat the system.

[0098] When the fluid-thermal coupling sensitivity index (Zero gain): This indicates that increased turbulence no longer leads to improved efficiency, reaching the "optimal impedance matching" point, i.e., the boiling critical point. In this embodiment, the fluid-thermal coupling sensitivity index converges to [value missing] within 120 seconds after entering the boiling stage. The zero value interval.

[0099] When the fluid-thermal coupling sensitivity index (Negative benefit): This indicates that increasing the heat output actually leads to a decrease in efficiency, possibly indicating "cavitation heat resistance" (excessive bubbles hindering heat transfer) or "precursors to burn the bottom." The control strategy should immediately implement a reduction in heat output. This represents a leap from "temperature-based control" to "energy efficiency-based control."

[0100] The following calculation logic is given for the fluid-thermal coupling sensitivity index to avoid numerical calculation divergence caused by steady-state physical conditions. Specifically, the following division operation process with denominator regularization mechanism is executed: calculate the time difference between the second state data (thermal efficiency) and the first state data (turbulence intensity) after hysteresis alignment.

[0101] Obtain the preset "minimum perturbation resolution threshold". Detect the absolute value of the hysteresis turbulence increment: if the absolute value of the hysteresis turbulence increment is greater than the minimum perturbation resolution threshold, directly perform a division operation. );

[0102] If the absolute value of the hysteresis turbulence increment is less than or equal to the minimum disturbance resolution threshold, the calculation result is forcibly set to zero, or a "signed pseudo-division" is used (only the thermal efficiency increment is retained and multiplied by a preset saturation constant). The original ratio obtained from the above steps is input to a smoothing filter, and the final fluid-thermal coupling sensitivity index is output. This logic ensures that even under extremely stable turbulent conditions, the control system can still obtain a bounded, physically meaningful input signal.

[0103] Further defining step S401, the dynamic extreme value search control process in step S401 includes a robust degradation strategy; the strategy is configured as follows:

[0104] Real-time monitoring of the signal-to-noise ratio of the first state data and the convergence stability of the fluid-thermal coupling sensitivity index;

[0105] When the signal-to-noise ratio is detected to be lower than a preset threshold or the fluid-thermal coupling sensitivity index exhibits non-convergent high-frequency oscillations, the output of the dynamic extreme value search controller is blocked.

[0106] It automatically switches to the preset open-loop power curve control logic and generates abnormal status signals that characterize sensor failure or instability.

[0107] Furthermore, in step S401, under normal operating conditions where the output of the dynamic extremum search controller is not masked, the dynamic extremum search control process is further defined as executing a variable step-size adaptive optimization process based on phase plane trajectory constraints, which includes:

[0108] A two-dimensional phase plane state space is constructed with the fluid-thermal coupling sensitivity index as the horizontal axis and the time differential component of the fluid-thermal coupling sensitivity index as the vertical axis.

[0109] Calculate the Euclidean distance norm of the current state point relative to the origin in the two-dimensional phase plane state space, and define this norm as the generalized tracking error potential energy.

[0110] A nonlinear adaptive decay function is constructed based on the generalized tracking error potential energy, and an adaptive gain modulation factor is generated using the adaptive decay function.

[0111] The adaptive gain modulation factor is used to modulate the preset basic gradient search direction to generate the final heating actuator power adjustment command.

[0112] Furthermore, in step S401, the step of calculating the generalized tracking error potential energy specifically includes:

[0113] Construct a two-dimensional state vector, where the first dimension component of the two-dimensional state vector is the current fluid-thermal coupling sensitivity index, and the second dimension component is the derivative of the fluid-thermal coupling sensitivity index with respect to time;

[0114] Calculate the Euclidean norm of the two-dimensional state vector, that is, calculate the sum of the squares of the first dimension component and the squares of the second dimension component, and perform a square root operation on the sum.

[0115] The result of the square root operation is defined as the generalized tracking error potential energy, which is used to characterize the comprehensive distance of the current state of the system from the ideal equilibrium point in the phase plane.

[0116] Furthermore, the steps for generating the adaptive gain modulation factor specifically include:

[0117] Read the minimum convergence step size benchmark and the maximum search gain limit from the preset external configuration file;

[0118] The generalized tracking error potential energy is nonlinearly mapped using the hyperbolic tangent function to obtain the normalized gain adjustment coefficient;

[0119] The gain adjustment coefficient is projected onto the interval formed by the minimum convergence step size reference and the maximum search gain upper limit, and the adaptive gain modulation factor is output so that when the generalized tracking error potential energy approaches zero, the control step size automatically decays to the minimum reference value according to an exponential law, thereby achieving smooth locking of the fluid-thermal coupling sensitivity index within the preset dead zone.

[0120] For the dynamic extremum search control and robustness adjustment in step S401: In this embodiment, step S401 executes variable gain phase plane damping control based on Lyapunov stability theory. This step uses the "fluid-thermal coupling sensitivity index" output from step S301 as the core error variable. By constructing a phase plane trajectory of "error - error change rate", the dynamic damping factor is calculated in real time, thereby achieving oscillation-free soft landing convergence with "rapid approach when far from the extremum point and automatic introduction of high damping lock-up near the extremum point". The execution process of this step is divided into two logical levels: "robustness pre-protection" and "phase plane adaptive optimization".

[0121] For the first level: robust degradation strategy (pre-protection logic); perform the following robust monitoring:

[0122] The signal-to-noise ratio (SNR) of the first state data in step S101 is calculated in real time. The current SNR is compared with a preset threshold (10dB, taken from the 1% quantile of 100 batches of normal historical data).

[0123] Real-time monitoring of the convergence stability of the "fluid-thermal coupling sensitivity index" is performed. If non-convergent, high-frequency, large-amplitude oscillations are detected in this index (indicating severe false boiling or gas resistance), it is considered a sign of severe false boiling or gas resistance.

[0124] When any of the above abnormal conditions are triggered, the state machine immediately disables the output of the subsequent dynamic extreme value search controller and automatically jumps to the safety degradation mode. In this mode, the preset safe power curve (30% of rated power) is directly looked up and output, and a state abnormality signal characterizing sensor failure or instability is generated until a reset is performed.

[0125] For the second level: variable step size adaptive optimization based on phase plane trajectory;

[0126] Under normal operating conditions where the degradation strategy is not triggered, the controller executes the following three-stage calculation process. Initial values ​​for all key parameters are read from the local configuration file via the data loading module.

[0127] Before proceeding with the calculation process, the core variable is defined: sensitivity tracking error ( ): Directly mapped from the fluid-thermal coupling sensitivity index output in step S301. Positive values ​​represent under-stirring, negative values ​​represent over-stirring / cavitation, and zero values ​​represent the best match.

[0128] Based on the fluid-thermal coupling sensitivity index, the differential ratio is smoothed using a smoothing function. The width of the smoothing filter window is a key parameter determining control stability. Since the differential operation between the "second-state data" (thermal efficiency) and the "first-state data" (turbulence intensity) amplifies high-frequency quantization noise, without smoothing, the directly output sensitivity index will exhibit severe spikes, leading to actuator malfunctions. This embodiment uses a "moving average filter" for smoothing. The window width of the moving average is set to M1. The value of this parameter M1 is equal to or greater than the "thermal hysteresis time constant" in step S301. The number of sampling points corresponding to “)”. In this embodiment, The sampling period is 1 second and calibrated to 15 seconds, so the value of M1 is set to 15. Aligning the smoothing window with the thermal hysteresis time is physically equivalent to examining the average marginal gain within the "complete thermal response cycle", thus mathematically eliminating spurious sensitivity fluctuations caused by local thermal inhomogeneity.

[0129] Error phase plane momentum ( ): Characterizing the sensitivity index of fluid-thermal coupling The rate of change of the trend is used to predict whether the system is about to cross an extreme point.

[0130] Generalized error potential energy norm ( ): Quantify the comprehensive distance of the current system state from the ideal convergence point (origin) in the phase plane, and use it as a benchmark for adjusting the control intensity.

[0131] Adaptive gain modulation factor ( ): Dynamic coefficients generated based on the potential energy norm for real-time scaling of gradient descent step size, ranging from [0,1].

[0132] Two-dimensional phase plane state space construction steps: Read the differential weighting coefficients from the configuration file ( ), S-shaped function slope factor ( Minimum hold step size ) and maximum burst step length ( Simultaneously, the sensitivity index at the current time k is received as the sensitivity tracking error. .

[0133] It should be noted that the differential weighting coefficients ( This parameter is used to balance the weights of "current error" and "error change trend" in potential energy calculation. In this embodiment, this parameter is set to 2.0 seconds. The physical dimension of this parameter is "time," and its numerical meaning lies in predicting how long the error will develop in the future. Setting it to 2.0 seconds is based on step S101, "the lowest cutoff frequency of fluid dynamics (0.5Hz, period 2 seconds)." This indicates that the evaluation scale of "momentum" is consistent with the physical period of fluid bubble generation, preventing over-prediction. Excessive boiling can cause the system to overreact to normal boiling pulsations, or it may be due to insufficient prediction. (Too small) leads to a lag in suppressing the "false boiling" phenomenon.

[0134] Slope factor ( This parameter determines the sensitivity of the controller when switching from "search state" to "lock state". It also defines the width of the "linear control region" near the origin of the phase plane. The larger the value, the steeper the tanh function near zero, and the more drastic the system's response to small errors. Deterministic computation logic: Define the "allowable residual boundary." That is, under the optimal matching state, the "generalized error potential norm" is allowed. The maximum thermal noise fluctuation value (denoted as) exists. This embodiment is based on historical data statistics, and this value is usually taken as 0.05. A gain attenuation target is set at this boundary. That is, when the error potential energy reaches... At this point, the adaptive gain is expected to decay to 50% of the maximum step size (meaning the tanh function value is 0.5).

[0135] According to the mathematical properties of the hyperbolic tangent function ( Perform a division operation: divide the constant 0.55 by the allowable residual boundary. (0.05), obtained The calculated value is 11.0. This parameter setting ensures that the controller responds at full speed when the system error is greater than the thermal noise floor; and once it enters the noise floor range, the control gain decays rapidly, avoiding the "high-frequency jitter" phenomenon near the optimal operating point.

[0136] Call the first-in-first-out (FIFO) queue to retrieve the sensitivity tracking error from the previous time step. The momentum of the error phase plane is calculated using a second-order backward difference algorithm: ;

[0137] In the two-dimensional phase plane, a preset static position weighting coefficient is introduced. With dynamic damping weight coefficient And the sum of the two is 1, and their initial values ​​are set to be equal; use these two coefficients to perform standardized weighting on the state vector, and calculate its weighted Euclidean norm:

[0138] ;in, The magnitude used to normalize the sensitivity index. This is used to map the time differential to the same order of magnitude as the sensitivity exponent. The calculation result... This is the generalized tracking error potential energy, which represents the comprehensive "energy" distance of the current state from the ideal equilibrium origin in the weighted phase plane.

[0139] By using the hyperbolic tangent function (tanh) to perform a nonlinear mapping on the generalized error potential energy, the normalized gain adjustment coefficient is obtained. : Specifically, in the dynamic extreme value search control of step S401, in order to achieve adaptive adjustment of the control step size, the following nonlinear mapping logic is executed to generate an adaptive gain modulation factor:

[0140] Obtain the generalized tracking error potential energy calculated at the current moment, and the preset slope factor ( The generalized tracking error potential energy is multiplied by a slope factor to obtain an intermediate weighting variable. The value of this slope factor determines the sensitivity of the control system near the extreme point. In this embodiment, it is configured such that when the error potential energy equals a preset thermal noise floor, the final output gain coefficient decays to 50% of the peak value. Hyperbolic tangent mapping is then performed. The intermediate weighting variable is input into the hyperbolic tangent (tanh) activation function. This function performs a nonlinear transformation and outputs a normalized gain adjustment coefficient ranging from 0 to 1. The physical meaning of this transformation is as follows: when the error potential energy is large, the output value approaches 1, allowing the control logic to maintain a large step search; when the error potential energy approaches zero, the output value decreases linearly and rapidly, achieving fine-tuning locking. This is achieved using a normalized gain adjustment coefficient. Linear interpolation is performed within the numerical range defined by the preset "minimum hold step size" and "maximum burst step size", and the calculated specific value is the final adaptive gain modulation factor.

[0141] when When moving away from the origin, ;when At that time, due to the high linearity of tanh near zero, the gain adjustment coefficient... It rapidly decays to 0. The final adaptive gain modulation factor is generated using interval projection logic. Even if it converges completely (error is 0), it still retains a small amount of... This is to maintain the ability to detect minute drifts and prevent control deadlock.

[0142] By combining the basic gradient direction and the adaptive step size, the power adjustment increment for the next cycle is synthesized. and update the total power command. : in, The rated power base, The sign function is (positive error increases power, negative error decreases power).

[0143] right Perform upper and lower limit truncation processing (limiting to the range of 0% to 100% of rated power) and send the final valid control command to the hardware driver layer (DAC interface).

[0144] Operating Condition Verification 1: Sudden Boiling (False Boiling) At this time, the number of bubbles increases rapidly, leading to sensitivity tracking error. The momentary negative change and the error phase plane momentum ( The momentum term increases. In the computational logic, a large momentum term leads to an increase in the generalized tracking error potential energy. The surge, in turn, causes the adaptive gain modulation factor to... Instantly approaching the maximum value It will rapidly reduce heating power in maximum increments, effectively suppressing the risk of overflowing.

[0145] Operating Condition Verification 2: Steady-State Fine-Tuning. When the liquid is in a stable, slightly boiling state, the sensitivity tracking error fluctuates slightly around 0. The calculated potential energy norm is small, and the adaptive gain modulation factor falls back to... The heating power is adjusted only slightly, avoiding frequent switching of the relay, extending the equipment's lifespan, and achieving a "soft landing" effect similar to high-damping mechanical systems.

[0146] The core control variable proposed in this invention is the fluid-thermal coupling sensitivity index ( ). The mathematical essence of this exponent is the eigenvalue of the partial derivative of the thermal efficiency with respect to the turbulence intensity after regularization. This exponent, together with the control logic based on the weighted phase plane, constitutes the core algorithm of this invention.

[0147] Under the combined effect of regularization (to prevent the denominator from being zero) and the weighted Euclidean norm, when the fluid-thermal coupling sensitivity index... The higher the value, the easier it is to reach the under-stirred state of the "high marginal gain region." In this state, the thermal convection resistance within the liquid solution more easily becomes dominant. That is, a small increase in physical turbulence can break down boundary layer thermal resistance, leading to a significant improvement in thermal efficiency. This indicates that most of the current input energy is effectively converted into convective heat transfer, without saturation. More heating power is needed to maximize the energy efficiency benefits in this region.

[0148] When the fluid-thermal coupling sensitivity index The closer it gets to zero (in this example) )hour:

[0149] The more easily the liquid reaches the "optimal impedance matching zone" (boiling point), the less effective the increase in thermal efficiency will be, according to the law of diminishing marginal utility. Physically, this means the convective heat transfer coefficient inside the liquid is more likely to reach its limit, and the more saturated the energy transfer channels are. Therefore, it is crucial to maintain the current power or make only minor adjustments, and to avoid ineffective mechanical energy dissipation (violent boiling) or the risk of overflow due to overheating.

[0150] When the fluid-thermal coupling sensitivity index As the temperature approaches its negative minimum, it becomes more prone to entering a "negative energy efficiency gain zone" (cavitation resistance / overheating state). In this state, increasing the heating power can actually lead to a decrease in thermal efficiency. Physically, this corresponds to an excessive buildup of bubbles on the heating surface (a precursor to film boiling), creating film thermal resistance, or the drug solution becoming more susceptible to localized overheating and degradation of the active ingredient. Therefore, a power reduction operation is increasingly necessary.

[0151] Furthermore, to ensure the robustness of the algorithm in industrial settings, the design of the following key parameters follows strict physical logic: thermal hysteresis time constant ( Alignment accuracy and The causal confidence levels are positively correlated. There is an inherent thermal inertia delay in the transmission of fluid disturbance (cause) to the temperature sensor (result). If no thermal inertia delay is introduced... Time-domain compensation results in a fluid-thermal coupling sensitivity index that is the ratio of "current thermal efficiency" to "turbulence at irrelevant moments," leading to spurious correlations. By aligning offline calibration with real-time hysteresis in step S301, the algorithm logic is ensured to conform to the physical law of "cause and effect," eliminating control oscillations.

[0152] Example 2:

[0153] This embodiment constructs a dynamic simulation scenario of the traditional Chinese medicine extraction process based on digital twin technology. This scenario aims to verify the modified "fluid-thermal coupling sensitivity index (FDI)". The response logic of the "weighted phase plane control algorithm" and its "weighted phase plane control algorithm" under different physical conditions. To ensure the rigor and reproducibility of the data, this embodiment uses the following calibrated model parameters: Physical object: 500L standard multi-functional extraction tank, equipped with an electric heating jacket with a maximum power of 10kW.

[0154] Thermal hysteresis time constant Seconds. Minimum perturbation resolution threshold. System rated power base kW. Static position weighting coefficient Dynamic damping weighting coefficient Slope factor of S-shaped function (Based on thermal noise floor) It is deduced that, i.e. Minimum hold step size Maximum search gain .

[0155] Real-time acquisition of turbulence increments after hysteresis alignment With thermal efficiency increment The fluid-thermal coupling sensitivity index is calculated using regularized logic. The error phase plane momentum, characterized by its rate of change over time, is calculated using second-order differences. Calculate the weighted generalized potential energy norm: .

[0156] Generate adaptive gain using hyperbolic tangent mapping: .

[0157] Combined with the sign function, output power regulation command Please refer to Table 1 below for details:

[0158] Table 1: Examples of fluid-thermodynamic coupling system response calculations under different physical conditions

[0159]

[0160] In scenario 2, the fluid-thermal coupling sensitivity exponent drops to +0.10. At this point, the weighted generalized potential energy norm... Calculation verification: Final gain .

[0161] It is worth noting that in scenario 2, the base gain based on the separate estimation of the fluid-thermal coupling sensitivity index is approximately 0.80; while the final adaptive gain recorded in the table is 0.81. This tiny numerical increment (+0.01) is not a calculation error, but a direct reflection of the differential momentum term in the phase plane control strategy of this invention. The specific calculation logic is as follows: When constructing the generalized tracking error potential energy, a weighted Euclidean norm synthesis of a two-dimensional vector is performed. Although the position component is 0.10, a sensitivity micro-component of -0.05 is detected simultaneously. According to the potential energy calculation logic, the sum of the squares of these two components, after square root operation, slightly increases the norm of the synthesized comprehensive potential energy to approximately 0.1031. Because the nonlinear mapping function (hyperbolic tangent) has a high gradient amplification characteristic in the transition region (slope factor 11.0), the aforementioned small potential energy increment is amplified by the mapping, resulting in a final output gain of 0.81 (i.e., a rounded value of 0.814). This data detail strongly demonstrates that the control system of the present invention possesses full-dimensional state perception capability. It not only adjusts based on the current static error but also incorporates the dynamic trend (momentum) into the decision-making process, thereby maintaining a more robust control tension than traditional proportional control (considering only the static error) when approaching the extreme point.

[0162] This data demonstrates that, when approaching the extreme point, the power is not directly cut off, but rather the step size is smoothly reduced from 0.99 to 0.81. This linear transition characteristic verifies... The reasonable parameter settings effectively avoid the oscillations near the critical point that occur in traditional switch control.

[0163] In scenario 3, the fluid-thermal coupling sensitivity index drops to (Entering the best-match dead zone). At this point... Calculation verification: Final gain The data demonstrates that the gain automatically decays to a low level (0.23) when entering the "dead zone". This "soft lock" mechanism maintains the ability to track small drifts while avoiding frequent actuator jitter caused by measurement noise, achieving a "close yet distant" steady-state control.

[0164] Scenario 3 and Scenario 5 have the exact same sensitivity index. Since the "position error" is the same, the output gain should be the same. Source of difference: A large negative rate of change exists in scenario 5. This indicates that the sensitivity is dropping rapidly and an overshoot is imminent.

[0165] Potential energy calculation: Because the passive potential energy term is increased to 0.10, the system output gain is 0.81, significantly higher than the 0.23 in scenario 3. The system maintains a high control gain (0.81) even when sensitivity drops rapidly. Due to... Continuing to apply positive power is physically equivalent to applying a "reverse support force," preventing the sensitivity index from dropping below zero too quickly due to inertia and causing underheating. This "position-velocity" coordinated mechanism achieves a "soft landing" in the dynamic process, effectively suppressing overshoot.

[0166] based on The numerical features identified two key thresholds: the convergence dead zone threshold ( The result is based on statistical analysis of the system's thermal noise floor. In this embodiment, based on statistical analysis of 100 batches of historical data, the random fluctuation variance of the thermal efficiency is: ,set up Principles, Determined Within this range, any fluctuation in sensitivity is considered a mathematical "zero point".

[0167] Negative return threshold ( (This is based on physical experimental calibration of the cavitation initiation point.) When the bubble coverage exceeds 30%, the heat transfer coefficient decreases sharply, corresponding to... It fell below -0.2. Considering the safety margin, we set... As the reversal point, -0.2 is the emergency cut-off point.

[0168] Zone 1: Energy efficiency improvement zone Its boundary is determined by the significance test of the marginal gain. When When the noise floor is greater than 0.05, it means that the energy input has a definite positive output. The corresponding operation is to execute gradient ascent control, with the heating power increment... To rapidly increase turbulence intensity in order to catch up with peak energy efficiency.

[0169] Interval 2: Best-match dead zone, i.e., the preset dead zone. Its boundary is jointly determined by system thermal noise and sensor measurement uncertainty. Within this range, fluctuations in the fluid-thermal coupling sensitivity index are considered random noise rather than physical trends, indicating that it is already in the stationary extreme region of the energy efficiency function. The corresponding operation is to activate the high-damping lock-in mode and the adaptive gain modulation factor. Attenuation to minimum value The heating power remains dynamically constant, only compensating for heat loss, to achieve steady-state fine-tuning.

[0170] Interval 3: Impedance mismatch / risk avoidance interval Its boundary is determined by the thermodynamic inverse inflection point. When the derivative turns negative, it indicates the entry into a nonlinear "cavitation resistance" or "superheating" physical state, violating the conventional laws of convective heat transfer. The corresponding action is to immediately implement negative gradient control or an open-loop degradation strategy to reduce or cut off the heating power until... Return to the non-negative range to prevent the liquid from charring or the equipment from being damaged.

[0171] Further: Figure 3 This is a schematic diagram illustrating the results of a numerical verification experiment conducted on the "fluid-thermal coupling sensitivity index" and its "dynamic extreme value search control" in one embodiment of the present invention. The diagram visually demonstrates the distribution of the discrete operating condition data listed in "Table 1" of the specification on the system control law curve, aiming to verify the control algorithm within the standardized mathematical model (…). The deterministic response characteristics under ( ). In the figure, the horizontal axis represents the "weighted generalized potential energy norm" calculated according to step S4, which comprehensively characterizes the distance and trend of the system deviating from the optimal matching point; the vertical axis represents the "adaptive gain modulation factor" output by the controller. The blue solid line in the figure represents the theoretical control law curve constructed based on the hyperbolic tangent function (Tanh), and the red scattered points (S1-S7) correspond to the calculated verification values ​​of each typical physical condition in Table 1.

[0172] like Figure 3 As shown, the numerical calculation results clearly reveal the three-stage control characteristics of the "variable step size adaptive optimization based on phase plane trajectory constraints" of this invention:

[0173] Best-match dead zone: corresponds to the green area in the lower left corner of the diagram. Verification points S3 (critical point) and S4 (steady state) precisely fall at the bottom of the curve, and the gain automatically decreases to 0.23 and 0.01 respectively. This mathematically confirms that when in steady state or with only minor thermal noise, it can automatically enter a "soft lock" state, effectively suppressing the actuator's invalid jitter.

[0174] Linear adjustment zone: corresponds to the yellow area in the middle of the figure. Verification points S2 (transition zone) and S5 (high damping) fall on the steep upward segment of the curve, with a gain output of 0.81 (medium-high gain). This confirms that whether approaching the extreme point (S2) or experiencing a rapid drop in sensitivity (S5), sufficient control tension and linear adjustment capability can be maintained through the calculation of weighted potential energy, achieving smooth transition and overshoot suppression in the dynamic process.

[0175] Saturation search region: corresponds to the red area on the right side of the diagram. Verification points S1 (under-stirring) and S7 (precursor to burn) fall within the flat region at the top of the curve, with the gain stabilizing at 0.99 (maximum value). This confirms that, under extreme operating conditions far from the extreme point, it can overcome the small step size limitation and respond quickly with the maximum power change rate, achieving rapid correction of energy efficiency status.

[0176] Figure 1 This illustrates the mapping relationship between the physical decoction equipment and the core control algorithm steps. The left side shows the decoction container entity with basic data acquisition and execution capabilities, while the right side shows the four-step closed-loop control logic flow electrically connected to it. Figure 1 As shown in the diagram, the core content of this technical architecture diagram covers four key steps:

[0177] Fluid dynamic feature extraction (corresponding to step S011): By collecting high-frequency pressure time-series signals inside the decoction container (device shown on the left), and using detrending and bandpass filtering techniques, the first state data (pressure fluctuation variance index) characterizing the current physical turbulence mixing intensity of the medicinal liquid is calculated and generated, thus completing the mapping from physical signals to the fluid dynamic feature space.

[0178] Thermodynamic energy state observation (corresponding to step S201): The input energy data and the temperature rise response of the liquid medicine acting on the heating jacket are acquired. Using an energy conservation state observer based on extended Kalman filter (EKF), the second state data characterizing the current instantaneous effective energy absorption efficiency of the liquid medicine is calculated in real time, thus realizing the observation of the invisible thermodynamic state.

[0179] Fluid-thermal coupling index (corresponding to step S301): The diagram shows that the system performs time-domain lag compensation alignment on the first and second state data mentioned above, and calculates the partial derivative characteristic value of the rate of change of the two, thereby generating the fluid-thermal coupling sensitivity index and establishing a causal quantitative relationship between physical disturbance and chemical endothermia.

[0180] Dynamic extreme value search control (corresponding to step S401): The fluid-thermal coupling sensitivity index is used as the controlled variable. The dynamic extreme value search controller based on phase plane trajectory constraints generates adjustment commands and feeds them back to the heating actuator of the cooking container. This drives the fluid-thermal coupling sensitivity index to converge to a preset dead zone near zero, thereby locking the optimal impedance matching point of turbulence-endothermic flow in the physical device.

[0181] It should be noted that all computational logic in this application employs regression analysis, including but not limited to machine learning algorithms, to deeply analyze the collected parameters and identify their natural trends and interrelationships. In all computational formulas of this application, the parameters in each formula undergo dimensionless processing within a consistent range to ensure that different physical quantities are compared on the same scale. Dimensionless processing techniques include, but are not limited to, Min-Max-Normalization and Z-Score standardization. To decouple the core algorithm of this invention from specific application strategies and to ensure the configurability and ease of debugging of the technical solution, in the specific implementation path of this invention, all configurable operating parameters are read through a standardized "configuration interface." The data source of this configuration interface is a "data storage module" (e.g., a non-transitory computer-readable storage medium, such as a configuration file, database entry, or cloud configuration service), which is configured to store configuration data in key-value pair format.

[0182] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A dynamic target optimization control method for the decoction process of a compound traditional Chinese medicine for veterinary use, characterized in that, The specific steps include: S101, Constructing the fluid dynamics feature space: Within a continuous time window, the pressure time-series signal inside the decoction container is acquired; the pressure time-series signal is subjected to detrending processing and frequency band filtering, and first state data characterizing the current physical turbulence mixing intensity of the medicinal liquid is calculated and generated. S201, constructing a thermodynamic energy observation space: The input energy data of the heating actuator acting on the decoction container and the temperature rise response data of the medicinal liquid are acquired; using a preset energy conservation state observer, based on the residual between the input energy data and the temperature rise response data, the second state data characterizing the current instantaneous effective energy absorption efficiency of the medicinal liquid is calculated in real time and generated. S301, a generative fluid-thermodynamic coupling fusion index: Perform time-domain hysteresis compensation alignment on the first state data and the second state data; calculate the partial derivative eigenvalue of the rate of change of the second state data relative to the rate of change of the first state data at the current moment, and define the partial derivative eigenvalue as the fluid-thermal coupling sensitivity index; S401, execute dynamic extreme value search control: The fluid-thermal coupling sensitivity index is input as a controlled variable to the dynamic extreme value search controller; the dynamic extreme value search controller generates control commands for adjusting the power of the heating actuator based on gradient descent logic, so as to drive the fluid-thermal coupling sensitivity index to converge to a preset dead zone near zero, thereby locking the optimal impedance matching point of turbulence-endothermic reaction of the liquid under the current operating conditions.

2. The dynamic target optimization control method for the decoction process of veterinary traditional Chinese medicine compound according to claim 1, characterized in that: In step S101, the first state data is the pressure fluctuation variance index, and the calculation and generation process includes: The pressure timing signal is extracted using a sliding window algorithm; The DC liquid level component and high-frequency electromagnetic noise component in the pressure time series signal are removed by using a bandpass filter, while retaining the bubble dynamics frequency band component that characterizes the bubble formation and rupture features. Calculate the statistical variance of the bubble dynamics frequency band components and quantify the statistical variance into the pressure fluctuation variance index.

3. The dynamic target optimization control method for the decoction process of veterinary traditional Chinese medicine compound according to claim 2, characterized in that: In step S201, the energy conservation state observer is constructed based on the extended Kalman filter algorithm; The second state data is the system thermal efficiency coefficient, and the calculation process includes: Establish a nonlinear thermodynamic equation of state that includes sensible heat terms, heat dissipation loss terms, and reaction endothermic terms; The input energy data is used as the control input, and the temperature rise response data is used as the observation output. Using the extended Kalman filter algorithm, the implicit state variables in the nonlinear thermodynamic equation of state are recursively estimated by minimizing the observation residuals, thereby obtaining the thermal efficiency coefficient of the system.

4. The dynamic target optimization control method for the decoction process of veterinary traditional Chinese medicine compound according to claim 3, characterized in that: In step S301, the fluid-thermal coupling sensitivity index characterizes the marginal gain of drug energy absorption efficiency relative to the intensity of physical turbulence. Its generation process includes: Access the preset hysteresis parameter table and read the thermal hysteresis time constant; Using the thermal hysteresis time constant, historical backtracking extraction is performed on the time series of the first state data to obtain the hysteresis first state data that has a causal correspondence with the current moment; Calculate the current rate of change of the second state data relative to time, and the historical rate of change of the first state data relative to time, respectively. Perform a division operation to calculate the ratio of the current rate of change to the historical rate of change, and perform moving average filtering on the ratio to determine the processed value as the fluid-thermal coupling sensitivity index.

5. The dynamic target optimization control method for the decoction process of veterinary traditional Chinese medicine compound according to claim 4, characterized in that: Step S401, the dynamic extreme value search control process, includes a robust degradation strategy; the strategy is configured as follows: Real-time monitoring of the signal-to-noise ratio of the first state data and the convergence stability of the fluid-thermal coupling sensitivity index; When the signal-to-noise ratio is detected to be lower than a preset threshold or the fluid-thermal coupling sensitivity index exhibits non-convergent high-frequency oscillations, the output of the dynamic extreme value search controller is blocked. It automatically switches to the preset open-loop power curve control logic and generates abnormal status signals that characterize sensor failure or instability.

6. The dynamic target optimization control method for the decoction process of veterinary traditional Chinese medicine compound according to claim 5, characterized in that: In step S401, under normal operating conditions where the output of the dynamic extremum search controller is not masked, the dynamic extremum search control process is further defined as executing a variable step-size adaptive optimization process based on phase plane trajectory constraints, which includes: A two-dimensional phase plane state space is constructed with the fluid-thermal coupling sensitivity index as the horizontal axis and the time differential component of the fluid-thermal coupling sensitivity index as the vertical axis. Calculate the Euclidean distance norm of the current state point relative to the origin in the two-dimensional phase plane state space, and define this norm as the generalized tracking error potential energy. A nonlinear adaptive decay function is constructed based on the generalized tracking error potential energy, and an adaptive gain modulation factor is generated using the adaptive decay function. The adaptive gain modulation factor is used to modulate the preset basic gradient search direction to generate the final heating actuator power adjustment command.

7. The dynamic target optimization control method for the decoction process of veterinary traditional Chinese medicine compound according to claim 6, characterized in that: In step S401, the step of calculating the generalized tracking error potential energy specifically includes: Construct a two-dimensional state vector, where the first dimension component of the two-dimensional state vector is the current fluid-thermal coupling sensitivity index, and the second dimension component is the derivative of the fluid-thermal coupling sensitivity index with respect to time; Calculate the Euclidean norm of the two-dimensional state vector, that is, calculate the sum of the squares of the first dimension component and the squares of the second dimension component, and perform a square root operation on the sum. The result of the square root operation is defined as the generalized tracking error potential energy, which is used to characterize the comprehensive distance of the current state of the system from the ideal equilibrium point in the phase plane.

8. The dynamic target optimization control method for the decoction process of veterinary traditional Chinese medicine compound according to claim 7, characterized in that: The specific steps for generating the adaptive gain modulation factor include: Read the minimum convergence step size benchmark and the maximum search gain limit from the preset external configuration file; The generalized tracking error potential energy is nonlinearly mapped using the hyperbolic tangent function to obtain the normalized gain adjustment coefficient; The gain adjustment coefficient is projected onto the interval formed by the minimum convergence step size reference and the maximum search gain upper limit, and the adaptive gain modulation factor is output so that when the generalized tracking error potential energy approaches zero, the control step size automatically decays to the minimum reference value according to an exponential law, thereby achieving smooth locking of the fluid-thermal coupling sensitivity index within the preset dead zone.

Images