An Adaptive Control Method for Drying Parameters Based on a Process Model

By decomposing path prediction bias into forward bias and lateral bias, and independently updating moisture and heat correction coefficients, the problem of inconsistent model parameters in the batch drying process of wet granulation of pharmaceuticals is solved, and the stability and consistency of drying control are achieved.

CN123114101APending Publication Date: 2026-11-17南京瑞邦制药设备有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611441510.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-17
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies struggle to effectively address batch-to-batch variations in raw materials and changes in heat and mass transfer states during wet granulation and batch drying of pharmaceuticals. This results in inconsistent model parameter updates, affecting the stability and consistency of inlet air temperature and volume settings.

Method used

By decomposing the path prediction bias into forward bias and lateral bias, which are used to update the moisture correction coefficient and the heat correction coefficient respectively, drying parameters are adjusted independently, forming an adaptive control method based on the process model, ensuring the independence and continuity of the moisture and heat correction processes.

Benefits of technology

It improves the targeting of drying control under batch fluctuation conditions and the continuity of setpoint output, reduces the cross-interference between moisture and heat changes, and improves the control consistency of pharmaceutical drying equipment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123114101A_ABST
    Figure CN123114101A_ABST
Patent Text Reader

Abstract

This invention relates to the field of pharmaceutical drying control technology, and discloses an adaptive control method for drying parameters based on a process model. The method includes: acquiring the material moisture content, inlet air temperature, and material temperature to form the actual drying movement direction; generating predicted moisture content and material temperature values ​​from the process model to form a path prediction deviation; decomposing the path prediction deviation into forward deviation and lateral deviation along the actual drying movement direction, ensuring that the forward deviation only enters the moisture correction branch to update the moisture correction coefficient, and the lateral deviation only enters the heat correction branch to update the heat correction coefficient; writing the updated moisture correction coefficient and heat correction coefficient back to the shared process model, and outputting the inlet air temperature setpoint and inlet air volume setpoint for the pharmaceutical drying equipment to execute. This invention is used for online control of the pharmaceutical drying process and can reduce cross-interference between the correction branches of the two models.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of pharmaceutical drying control technology, and more specifically, to an adaptive control method for drying parameters based on a process model. Background Technology

[0002] Batch drying of pharmaceuticals after wet granulation typically maintains the heat and mass transfer process within the drying chamber through inlet heating, air supply, and exhaust. The on-site control system adjusts the heater and blower based on the material moisture content, inlet air temperature, material temperature, and inlet air volume. Since batch raw materials, loading amounts, particle sizes, ambient humidity, and equipment thermal conditions fluctuate, and the rate of decrease in material moisture content interacts with changes in material temperature, a single fixed formula or model parameter is insufficient to cover the entire batch.

[0003] Current engineering practices often employ staged formulation, PID inner loop, or upper-level control based on process models. Staged formulation relies on preset timing to switch inlet air temperature and volume. PID inner loop primarily ensures that the executed quantity tracks the setpoint, but its ability to compensate for batch-to-batch raw material differences and changes in heat and mass transfer states is limited. Process model-based control can update the model and calculate subsequent setpoints using online measurements. However, when the measured material moisture content and temperature are inconsistent with the corresponding predicted results, existing online correction methods can easily alter multiple model parameters simultaneously, and the effects of moisture and heat changes on the correction results may be cumulative.

[0004] This hybrid update can easily lead to mismatched error sources for the moisture correction coefficient and the heat correction coefficient, causing model parameters to interfere with each other in adjacent sampling periods. This, in turn, affects the continuity of the inlet air temperature setpoint and the inlet air volume setpoint, as well as the equipment response. In engineering practice, it is necessary to improve the differentiation between moisture and heat changes in online model correction, so that the updated process model can continuously provide executable and consistent setpoints for the heater and blower. Summary of the Invention

[0005] This invention provides an adaptive control method for drying parameters based on a process model, which solves the technical problems mentioned in the background art.

[0006] To achieve the above-mentioned objectives, the present invention provides the following technical solution.

[0007] An adaptive control method for drying parameters based on a process model includes:

[0008] The material moisture content, inlet air temperature, and material temperature during the drying process are obtained, and the actual drying movement direction is determined based on the material moisture content, inlet air temperature, and material temperature.

[0009] The process model generates predicted values ​​for moisture content and material temperature, and the path prediction deviation is formed based on the predicted values ​​for moisture content, material temperature, material moisture content, inlet air temperature, and material temperature.

[0010] Based on the actual drying movement direction, the path prediction deviation is decomposed into forward deviation and lateral deviation;

[0011] Set moisture correction coefficients and heat correction coefficients so that forward bias only enters the moisture correction branch to update the moisture correction coefficient, and lateral bias only enters the heat correction branch to update the heat correction coefficient. Path prediction bias is not used as a common update input for the moisture correction branch and the heat correction branch before decomposition.

[0012] Write the updated moisture and heat correction factors back into the shared process model;

[0013] Based on the shared process model, the executable inlet air temperature setpoint and inlet air volume setpoint for the drug drying equipment are output.

[0014] Compared with the prior art, the present invention has the following substantial features and significant progress:

[0015] Compared to control methods that use the same path prediction deviation for multiple model correction branches, this invention determines the actual drying movement direction based on material moisture content, inlet air temperature, and material temperature. It decomposes the path prediction deviation into forward and lateral deviations, ensuring that the forward deviation is used only to update the moisture correction coefficient and the lateral deviation is used only to update the heat correction coefficient. This reduces cross-interference between moisture and heat changes during model correction, ensuring that the updated objects of the two correction coefficients are consistent with the corresponding process changes. After the update results are written back to the shared process model, they can be directly used to generate the inlet air temperature and inlet air volume setpoints for the pharmaceutical drying equipment. This improves the targeting of model correction under batch fluctuation conditions and enhances the continuity of setpoint output and the consistency of drying control. Attached Figure Description

[0016] Figure 1 This is a control module diagram of the present invention;

[0017] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0018] The following description is provided in conjunction with the accompanying drawings. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the invention; those skilled in the art can make equivalent substitutions or combinations for the specific implementations.

[0019] The following combination Figure 1 and Figure 2The present invention will be further described below. The pharmaceutical drying equipment referred to in this embodiment refers to batch drying of solid pharmaceutical wet granules using controlled air intake, capable of obtaining material moisture content, air intake temperature, and material temperature in each drying control cycle, and capable of adjusting the air intake temperature and air intake volume separately. This embodiment is applicable to solid oral dosage form wet granules with batch loading of 20 kg to 120 kg, wet particle size of 0.2 mm to 1.5 mm, and material moisture content of 2.0% to 18.0%. The pharmaceutical drying equipment 111 includes a drying chamber, an air intake heater, a blower, and actuator inner rings for the air intake heater and the blower. After the material to be dried is loaded into the drying chamber, the air intake heater adjusts the air intake temperature, and the blower adjusts the air intake volume. The airflow contacts the material and completes heat and mass transfer. The industrial controller obtains the material moisture content, air intake temperature, and material temperature in each drying control cycle, and outputs the air intake temperature setpoint and air intake volume setpoint to the pharmaceutical drying equipment 111.

[0020] For different batch loading amounts and wet particle sizes within the aforementioned applicable scope, the industrial controller maintains the data relationships between the actual drying movement direction, path prediction deviation, forward deviation, lateral deviation, moisture correction branch 116, heat correction branch 117, and shared process model 118 unchanged, and determines the fixed model parameters according to the calibration process described in this embodiment. The formal batch uses the same drying control cycle as when calibrating the fixed model parameters; when the drying control cycle is changed, the fixed model parameter calibration and adaptive parameter playback are re-executed.

[0021] See Figure 1 The industrial controller includes a drying process data acquisition module 112, an actual drying movement direction formation module 113, a path prediction deviation formation module 114, a path prediction deviation decomposition module 115, a moisture correction branch 116, a heat correction branch 117, a shared process model 118, and a control parameter output module 119. The drying process data acquisition module 112 receives the material moisture content, inlet air temperature, and material temperature. The actual drying movement direction formation module 113 forms the actual drying movement direction based on the above data. The path prediction deviation formation module 114 receives the predicted moisture content and material temperature generated by the shared process model 118 and forms a path prediction deviation within the same path space. The path prediction deviation decomposition module 115 decomposes the path prediction deviation into forward deviation and lateral deviation according to the actual drying movement direction.

[0022] Moisture correction branch 116 and heat correction branch 117 are each provided with independent input ports. The path prediction deviation decomposition module 115 sends the forward deviation to the independent input port of moisture correction branch 116 and the lateral deviation to the independent input port of heat correction branch 117. Moisture correction branch 116 updates its moisture correction coefficient accordingly, and heat correction branch 117 updates its heat correction coefficient accordingly. The updated moisture and heat correction coefficients are written back to the shared process model 118, and the control parameter output module 119 then generates the inlet air temperature setpoint and inlet air volume setpoint based on the shared process model 118. This port mapping ensures that the path prediction deviation before decomposition does not constitute the common update input for the two correction branches.

[0023] In this embodiment, the process model referred to in the claims is a shared process model 118. The shared process model 118, the moisture correction branch 116, the heat correction branch 117, and the control parameter output module 119 are all housed within the same industrial controller and are sequentially invoked by the periodic tasks triggered by the drying control cycle. The moisture correction coefficient and the heat correction coefficient are stored as two online model parameters of the shared process model 118 in the same parameter area; both the path prediction deviation formation module 114 and the control parameter output module 119 read the current values ​​from this parameter area.

[0024] In this embodiment, the material moisture content is expressed as a percentage of wet basis mass corresponding to the loss on drying method, and the percentage point value is used in the program calculation; the inlet air temperature and material temperature are expressed in degrees Celsius; the inlet air volume is expressed as the operating volumetric flow rate of the pharmaceutical drying equipment 111 under the current inlet air temperature and pipeline pressure, expressed in cubic meters per hour. The predicted moisture content value and the material moisture content use the same benchmark, and the predicted material temperature value and the material temperature use the same temperature scale. The above data definitions remain unchanged within the same batch.

[0025] The material moisture content is obtained using an online moisture analyzer. The measuring window of the online moisture analyzer is located on the side wall of the drying chamber, with its center positioned above the airflow distribution plate at approximately 55% of the height of the stationary material layer. The end face of the measuring window is flush with the inner wall of the drying chamber. The material temperature sensor is positioned at the same height as the measuring window, with its measuring end extending radially into the drying chamber to a position no more than 10% of the inner diameter of the drying chamber's central axis. The inlet air temperature sensor is located downstream of the inlet air heater at least five times the inner diameter of the inlet air duct and upstream of the air inlet of the drying chamber at least three times the inner diameter of the inlet air duct. The measuring end is located at the center of the inlet air duct's cross-section. The upstream straight pipe section of the inlet air volume measuring device is at least ten times the inner diameter of the inlet air duct, and the downstream straight pipe section is at least five times the inner diameter of the inlet air duct. The response time of each measuring device is no more than 10 seconds. The measurement results are converted by the industrial controller to generate the material moisture content, inlet air temperature, material temperature, and inlet air volume.

[0026] The calibration of the online moisture analyzer was performed simultaneously with the fixed model parameter calibration. At the initial, intermediate, and near-target moisture content stages of each calibration batch, three material samples were continuously obtained from sampling ports adjacent to the measurement window. The industrial controller recorded a millisecond-level timestamp when each material sample left the drying chamber, and the arithmetic mean of the online moisture analyzer readings within 15 seconds before and after this timestamp was used as the corresponding online reading. The three material samples were dried to constant weight according to the loss-on-drying conditions specified in the pharmacopoeia or production procedures. The wet basis mass percentage was obtained based on the initial mass, and the arithmetic mean of the three wet basis mass percentages was taken as the corresponding loss-on-drying detection result.

[0027] Using online readings from six calibration batches as independent variables and corresponding loss-on-drying (LOD) test results as dependent variables, the first-order linear calibration relationship was determined using ordinary least squares (OLS). The coefficient of determination for the first-order linear calibration relationship was no less than 0.98, and the root mean square error of the moisture content of the calibrated material did not exceed 0.30 percentage points, with a maximum absolute error not exceeding 0.60 percentage points. The first-order linear calibration relationship meeting the above conditions was used in the industrial controller for the official batches. After the online moisture analyzer underwent recalibration, measurement window maintenance, or material formulation changes, the first-order linear calibration relationship was re-determined using the same sampling, pairing, and calculation process.

[0028] Before the start of a pharmaceutical drying batch, the industrial controller is loaded with the target moisture content, allowable moisture content difference, upper limit of material temperature, lower limit of inlet air temperature, upper limit of inlet air temperature, lower limit of inlet air volume, upper limit of inlet air volume, upper limit of single temperature change, upper limit of single air volume change, drying control cycle, and prediction steps. The target moisture content and allowable moisture content difference determine the range of material moisture content at the end of the batch, and the upper limit of material temperature limits the material temperature. The lower limit of inlet air temperature, upper limit of inlet air temperature, lower limit of inlet air volume, and upper limit of inlet air volume limit the range of candidate values ​​for the control parameter output module 119, and the upper limit of single temperature change and upper limit of single air volume change limit the change amplitude of adjacent control steps. The drying control cycle determines the call interval of the periodic task, and the prediction steps determine the number of control steps included in the future prediction time domain.

[0029] The dimensions of the moisture content weight are negative square percentage points, the temperature action weight is negative square degrees Celsius, and the air volume action weight is negative square (cubic meters per hour), ensuring that all three weighted square terms in the objective function are dimensionless. The three weights are determined through time-domain playback of future predictions for the calibration batch. During playback, the same material type, batch load, drying control cycle, and control constraints as the official batch are used. The moisture content weight is adjusted according to the control tracking priority from largest to smallest, and the temperature and air volume action weights are adjusted according to the allowable operating frequencies of the inlet heater and blower. The combination that meets the upper limit of material temperature and has a root mean square error of no more than 0.30 percentage points in the validation data is selected; if multiple combinations exist, the combination with the smaller total change in adjacent control steps for the inlet temperature setpoint and inlet air volume setpoint is selected.

[0030] The shared process model 118 includes a moisture decline coefficient, a temperature-induced drying coefficient, an airflow-induced drying coefficient, a temperature difference-induced heating coefficient, an airflow-induced heating coefficient, and an evaporative cooling coefficient. The moisture decline coefficient is dimensionless; the temperature-induced drying coefficient is in percentage points per degree Celsius; the airflow-induced drying coefficient is in percentage points divided by cubic meters per hour; the temperature difference-induced heating coefficient is dimensionless; the airflow-induced heating coefficient is in degrees Celsius divided by cubic meters per hour; and the evaporative cooling coefficient is in degrees Celsius per percentage point. Both the moisture correction coefficient and the heat correction coefficient are dimensionless.

[0031] The fixed model parameters were determined using six calibration batches of the pharmaceutical drying equipment 111. All six calibration batches used the same material formulation, with a batch load of 60 kg and an allowable deviation of no more than 1.2 kg. The initial material moisture content was 10.5% to 11.5%, the initial material temperature was 23°C to 27°C, and the drying control cycle was 30 seconds. The first four calibration batches were used for parameter identification. In the first calibration batch, the airflow rate was maintained at 700 cubic meters per hour, and the airflow temperature was successively maintained at 40°C, 50°C, 60°C, and 70°C. The second calibration batch used the same airflow temperature sequence and maintained the airflow rate at 1300 cubic meters per hour. Each airflow temperature holding phase lasted 10 minutes, and adjacent phases transitioned sequentially according to the upper limit of a single temperature change within each drying control cycle.

[0032] The third calibration batch maintained the inlet air temperature at 50 degrees Celsius, and successively maintained the inlet air volume at 700 cubic meters per hour, 900 cubic meters per hour, 1100 cubic meters per hour, and 1300 cubic meters per hour. The fourth calibration batch used the same inlet air volume sequence and maintained the inlet air temperature at 65 degrees Celsius. Each inlet air volume maintenance phase lasted for 10 minutes, and adjacent phases transitioned sequentially according to the upper limit of the single air volume change in each drying control cycle. In each of the four calibration batches, the material moisture content, inlet air temperature, material temperature, and inlet air volume were recorded synchronously in each drying control cycle, and data pairs for adjacent drying control cycles were formed according to millisecond-level timestamps.

[0033] During parameter identification, the moisture correction factor and heat correction factor were fixed at 1.00, and the data were arranged in chronological order according to the four calibration batches and timestamps. First, the difference between the predicted moisture content and the material moisture content was used as the moisture residual. The Lawson-Hanson active set non-negative least squares method was used to determine the moisture reduction coefficient, temperature-induced drying coefficient, and airflow-induced drying coefficient. Then, using the obtained predicted moisture content, the difference between the predicted material temperature and the material temperature was used as the temperature residual. The same method was used to determine the temperature difference heating coefficient, airflow-induced heating coefficient, and evaporative cooling coefficient. The moisture residual was divided by 0.30 percentage points, and the temperature residual was divided by 1.0 degree Celsius. Both types of normalized residuals were given the same weight in the sum of squared residuals. The initial values ​​of the six fixed model parameters are all set to zero. The moisture reduction coefficient, temperature drying coefficient, and air volume drying coefficient are determined in sequence, and the temperature difference heating coefficient, air volume heating coefficient, and evaporation cooling coefficient are determined to form one round of iteration. When the relative change of the sum of squares of the residuals between two adjacent rounds is less than one part per million, the result of this round is taken. If this condition is not met in advance, the result of the 200th round is taken.

[0034] The fifth and sixth calibration batches were used for independent validation. The fifth calibration batch used combinations of inlet air temperature of 45°C and inlet air volume of 800 m³ / h, 55°C and inlet air volume of 1000 m³ / h, and 65°C and inlet air volume of 1200 m³ / h, respectively. The sixth calibration batch used the above three combinations in reverse order. Each combination was maintained for 10 minutes, with transitions between combinations based on the upper limit of a single temperature change and the upper limit of a single air volume change. The two validation batches were substituted into the shared process model 118 in the order of sampling time. When the root mean square error of the moisture content prediction value did not exceed 0.30 percentage points and the maximum absolute error did not exceed 0.60 percentage points, and the root mean square error of the material temperature prediction value did not exceed 1.0°C and the maximum absolute error did not exceed 2.0°C, the fixed model parameters were used. After changing the material formula, the batch loading amount changes by more than 10% relative to the calibration batch, the drying control cycle is changed, the online moisture analysis device is recalibrated, or the air inlet heater and blower are maintained to affect the output capacity, the fixed model parameters are re-determined according to the same calibration process.

[0035] The industrial controller sets initial values ​​for moisture and heat correction coefficients and loads lower and upper limits for moisture and heat correction, as well as moisture update step size, heat update step size, zero-prevention amount, and minimum path movement. The moisture and heat correction coefficients and their limits, moisture and heat update step sizes, zero-prevention amount, and minimum path movement are all dimensionless. The moisture and heat update step sizes control the update amplitude within a drying control cycle. The zero-prevention amount has the same dimension as the square of the direction sensitivity. The minimum path movement is used to determine whether the path movement is sufficient to update the actual drying movement direction.

[0036] The adaptive parameters described above were implemented using sequential data playback across two validation batches. The initial values ​​for both the moisture correction coefficient and the heat correction coefficient were 1.00, with upper and lower limits set to 0.50 and 1.50 respectively. The moisture update step size and the heat update step size were selected from 0.05, 0.10, 0.15, and 0.20, respectively; the zero-prevention amount was selected from 0.0001, 0.001, and 0.01; and the minimum path movement was selected from 0.002, 0.005, and 0.010. The shared process model 118 prediction and two correction branch updates were performed according to the sampling order. Combinations with a root mean square error (RMSE) of moisture content prediction not exceeding 0.30 percentage points, a RMSE of material temperature prediction not exceeding 1.0 degree Celsius, and single-cycle changes in both the moisture correction coefficient and the heat correction coefficient not exceeding 0.05 were selected. When multiple combinations met the conditions, the combination with the smaller verification error in moisture content and material temperature prediction was used.

[0037] See Figure 2 In step S211, the drying process data acquisition module 112 acquires the material moisture content, inlet air temperature, and material temperature, and also acquires the inlet air volume required by the shared process model 118. The industrial controller's periodic task sends a unified sampling trigger signal every 30 seconds, and each measurement channel writes a 32-bit floating-point engineering quantity and a millisecond-level timestamp into the batch data area. Using the sampling trigger time as the alignment reference, data closest to that time and with a time difference not exceeding 1 second is selected from each measurement channel to form the material moisture content, inlet air temperature, material temperature, and inlet air volume for the same drying control cycle. The data for the current drying control cycle, the data for the previous drying control cycle, and the corresponding prediction results generated by the shared process model 118 are stored using the same cycle number.

[0038] At the start of a batch, the industrial controller initiates a cycle task after the inlet air heater and blower reach their initial operating state. The first sampling point acquired by the cycle task simultaneously satisfies the conditions that the material moisture content is greater than zero and the inlet air temperature is higher than the material temperature. The actual drying movement direction forming module 113 stores the initial moisture content and initial thermal drive difference corresponding to this first sampling point, and the initial moisture content and initial thermal drive difference remain unchanged within the same batch. The corresponding relationship is as follows:

[0039]

[0040] In the formula, The initial moisture content, For the first The moisture content of the material at each sampling time; at the first sampling point, Write the initial moisture content to the storage unit. The initial moisture content is a percentage value and is greater than zero, which will be used for dimensionless processing of subsequent measured and predicted path points in this batch.

[0041]

[0042] In the formula, For the initial thermally driven difference, For the first The inlet air temperature at each sampling time, For the first The material temperature at each sampling time. The initial thermal drive difference, expressed in degrees Celsius and greater than zero, serves as the dimensionless reference for the thermal drive difference. The first sampling point also serves as the starting point for recording subsequent path movement.

[0043] Step S213 generates predicted moisture content and material temperature values ​​based on shared process model 118. Shared process model 118 uses the material moisture content and material temperature from the previous sampling time as its state, and the actual inlet air temperature and inlet air volume of the corresponding drying control cycle as its input. A moisture correction coefficient applies to the change in material moisture content, a heat correction coefficient applies to the change in material temperature, and an evaporation cooling coefficient reflects the impact of the decrease in material moisture content on the change in material temperature. Shared process model 118 generates the one-step prediction result for the current sampling time according to the following relationship:

[0044]

[0045] In the formula, According to the first The data generated at the nth sampling time is the first Predicted moisture content at each sampling time, in percentage points; For the first Moisture content of the material at each sampling time, in percentage points; This is the current moisture correction factor, dimensionless. The moisture reduction coefficient is dimensionless. This is the temperature-induced drying coefficient, expressed as percentage points per degree Celsius. This is the air volume drying coefficient, expressed as percentage points divided by cubic meters per hour; For the first The inlet air temperature at each sampling time, The lower limit of the intake air temperature; For the first Air intake at each sampling time, This represents the lower limit of the air intake volume. All three items within square brackets, as well as the predicted moisture content, are expressed as percentage points.

[0046]

[0047] In the formula, According to the first The data generated at the nth sampling time is the first The predicted material temperature at each sampling time, in degrees Celsius; For the first The material temperature at each sampling time; This is the current heat correction factor, dimensionless; The coefficient of temperature rise due to temperature difference is dimensionless. This is the air volume temperature rise coefficient, expressed in degrees Celsius divided by cubic meters per hour; This is the evaporative cooling coefficient, expressed in degrees Celsius per percentage point. Intake air temperature, For air intake volume, This is the lower limit of the air intake volume. The moisture content of the material. This represents the predicted moisture content. All terms in this formula are expressed in degrees Celsius.

[0048] Step S212 involves the actual drying movement direction being formed by the actual drying movement direction forming module 113. The actual drying movement direction forming module 113, according to the... The thermal drive difference is determined by the inlet air temperature and material temperature at each sampling time. Then, the measured path point is formed using the material moisture content and thermal drive difference at the same sampling time. The thermal drive difference is expressed in degrees Celsius, and its calculation relationship is as follows:

[0049]

[0050] In the formula, For the first The thermal difference at each sampling time For the first The inlet air temperature at each sampling time, For the first The material temperature at each sampling time. Since the dimensions and orders of magnitude of the material moisture content and thermal drive difference are different, the initial moisture content and initial thermal drive difference are used for dimensionless processing to form the measured path points:

[0051]

[0052] In the formula, For the first The measured path points at each sampling time. For the first Moisture content of the material at each sampling time. The initial moisture content, For the first The thermal difference at each sampling time The initial thermally driven difference is represented by the first component of the measured path point, which reflects the change in material moisture content relative to the initial moisture content, while the second component reflects the change in thermally driven difference relative to the initial thermally driven difference. Using the first sampling point of the same batch as a dimensionless reference allows for comparison of the drying state at consecutive sampling times within the same path space.

[0053] The actual drying movement direction forming module 113 stores the measured path points at adjacent sampling times and determines the path forward vector based on the path movement amount of the measured path points at adjacent sampling times. The path movement amount is:

[0054]

[0055] In the formula, For the first Path movement at each sampling time, This is the current measured path point. This represents the previous measured path point. The path movement includes changes in both material moisture content and thermal drive difference within a drying control cycle, and is used to characterize the actual movement of the drug drying process within the path space.

[0056]

[0057] In the formula, For the first The path forward vector at each sampling time. This is the Euclidean norm of the path movement. When the Euclidean norm of the path movement is greater than the minimum path movement, the actual drying movement direction forming module 113 performs normalization processing. The path forward vector is a dimensionless unit vector, and its two components correspond to the material moisture content direction and the thermal drive difference direction, respectively.

[0058] When the Euclidean norm of the path movement is not greater than the minimum path movement, the actual drying movement direction forming module 113 uses the path forward vector and path lateral quantity already generated in the previous drying control cycle. Since the first sampling point has not yet formed the previous measured path point, the industrial controller completes measurement and prediction in this drying control cycle and maintains the moisture correction coefficient and heat correction coefficient; path movement judgment is executed starting from the drying control cycle that obtains the second measured path point. This direction maintenance method does not change the input correspondence between the two correction branches.

[0059]

[0060] In the formula, For the first The horizontal path quantity at each sampling time. and These are the first and second components of the path forward vector, respectively. The path lateral quantity, obtained by arranging these components, is perpendicular to the path forward vector. The actual drying movement direction, jointly represented by the path forward vector and the path lateral quantity, provides a directional reference for subsequently decomposing the path prediction deviation into forward and lateral deviations.

[0061] Step S214 involves the path prediction deviation forming module 114 generating the path prediction deviation. The path prediction deviation forming module 114 generates the path prediction deviation based on the... The predicted thermal drive difference is formed by the predicted inlet air temperature and material temperature at each sampling time:

[0062]

[0063] In the formula, To predict the thermal driving difference, This is the current intake air temperature. The predicted material temperature is used. The predicted thermal drive difference and the thermal drive difference use the same inlet air temperature, so that the difference between the two is determined by the difference between the material temperature and the predicted material temperature.

[0064]

[0065] In the formula, To predict path points, This is the predicted moisture content. The initial moisture content, To predict the thermal driving difference, The initial thermal drive difference is used. The predicted path points and the measured path points are dimensionlessly processed using the same initial moisture content and initial thermal drive difference, so the corresponding components of the two path points can be directly compared.

[0066]

[0067] In the formula, For the first Path prediction bias at each sampling time These are the measured path points. The path prediction deviation is calculated using a path prediction deviation module 114. The first component of the path prediction deviation originates from the difference between the material moisture content and the predicted moisture content, while the second component originates from the difference between the thermally driven difference and the predicted thermally driven difference. The path prediction deviation generation module 114 then sends the path prediction deviation to the path prediction deviation decomposition module 115.

[0068] Path prediction bias decomposition and two corrected branch updates

[0069] Step S215 involves the path prediction deviation decomposition module 115 decomposing the path prediction deviation into forward deviation and lateral deviation. The path prediction deviation decomposition module 115 simultaneously receives the path prediction deviation, the forward vector, and the lateral quantity, and performs projections along the forward vector and lateral quantity respectively.

[0070]

[0071] In the formula, For the first Forward bias at each sampling time, This is the transpose of the path prediction bias. The forward deviation represents the difference between the predicted path point and the measured path point along the actual drying direction. Since the actual drying direction is determined by continuous measured path points, the forward deviation corresponds to the change in material moisture content along the existing drying trajectory during the drug drying process.

[0072]

[0073] In the formula, For the first Lateral deviation at each sampling time, The lateral deviation represents the degree to which the predicted path point deviates from the actual drying movement direction relative to the measured path point, and corresponds to the prediction difference in the thermally driven difference direction of the shared process model 118. The forward deviation and lateral deviation are obtained from the same path prediction deviation in two mutually perpendicular directions, and are used as update inputs for the two correction branches, respectively.

[0074] Steps S216 and S217 are executed sequentially within the same drying control cycle. The path prediction deviation decomposition module 115 writes the forward deviation to the independent input port of the moisture correction branch 116 and the lateral deviation to the independent input port of the heat correction branch 117; the moisture correction branch 116 and the heat correction branch 117 read the data from their respective independent input ports and complete the corresponding updates. The path prediction deviation before decomposition is only stored in the input data area of ​​the path prediction deviation decomposition module 115.

[0075] This implementation uses the explicit state change relationship of the shared process model 118 to determine the moisture direction sensitivity and the heat direction sensitivity. First, the baseline moisture change and the baseline temperature change are calculated:

[0076]

[0077] In the formula, For the first The baseline change in moisture content at each sampling time, expressed as a percentage point; is a dimensionless moisture reduction coefficient; The moisture content of the material at the previous sampling time; Temperature drying coefficient, expressed as percentage points per degree Celsius; The inlet air temperature at the previous sampling time; The lower limit of the intake air temperature; The drying coefficient is expressed as a percentage point divided by the air volume per cubic meter per hour. This represents the air intake volume at the previous sampling time. This is the lower limit of the air intake volume.

[0078]

[0079] In the formula, For the first The baseline temperature change at each sampling time, in degrees Celsius; It is a dimensionless coefficient of thermal rise due to temperature difference; The inlet air temperature at the previous sampling time; The material temperature at the previous sampling time; The temperature rise coefficient is expressed as the air volume per cubic meter per hour, measured in degrees Celsius. This represents the air intake volume at the previous sampling time. This is the lower limit of the air intake volume.

[0080] Based on the baseline values ​​of moisture change, temperature change, initial moisture content, initial thermal drive difference, and evaporation cooling coefficient, determine the relationship between the predicted path points and their changes in the path space when the moisture correction coefficient and heat correction coefficient change.

[0081]

[0082] In the formula, Sensitivity to water pathways, This is the baseline for moisture variation. The initial moisture content, The coefficient of performance is the evaporative cooling rate. This represents the initial thermally driven difference. The first component of the moisture path sensitivity indicates the change in the first component of the predicted path point corresponding to the predicted moisture content when the moisture correction coefficient changes. The second component indicates the change in the second component of the predicted path point corresponding to the change in the moisture content of the same material after evaporation and cooling.

[0083]

[0084] In the formula, For heat path sensitivity, The basic quantity is the change in temperature. This represents the initial thermally driven difference. The heat correction factor applies to the temperature change of the material, but not directly to the predicted moisture content. Therefore, the first component of the heat path sensitivity is zero, and the second component represents the change in the predicted thermally driven difference when the heat correction factor changes.

[0085]

[0086] In the formula, Sensitivity to water direction. This is the transpose of the path forward vector. This is the water path sensitivity. Water direction sensitivity characterizes the relationship between the predicted path point and the forward vector when the water correction coefficient changes. Based on this, water correction branch 116 converts the forward deviation into the update amount of the water correction coefficient.

[0087]

[0088] In the formula, Sensitivity to heat direction This is the transpose of the horizontal dimension of the path. This refers to the heat path sensitivity. Heat direction sensitivity characterizes the change in the lateral amount along the path of the predicted path point when the heat correction coefficient changes. The heat correction branch 117 uses this to convert the lateral deviation into an update amount of the heat correction coefficient. Both moisture direction sensitivity and heat direction sensitivity are determined by the state change relationship of the current drying control cycle in the shared process model 118.

[0089] Moisture correction branch 116 updates the moisture correction coefficient according to forward bias and moisture direction sensitivity:

[0090]

[0091] In the formula, This is the updated moisture correction factor. The moisture correction factor is as shown in the previous version. For water renewal step size, Forward deviation, Sensitivity to water direction. To prevent zero quantity, This is the lower limit for moisture correction. This represents the upper limit for moisture correction. All parameters mentioned above are dimensionless. The calculation is limited by the lower and upper limits of moisture correction: if the result is lower than the lower limit, the lower limit is used; if it is higher than the upper limit, the upper limit is used; if it is in between, the result is retained.

[0092] Heat correction branch 117 updates the heat correction coefficient according to lateral deviation and heat direction sensitivity:

[0093]

[0094] In the formula, For the updated heat correction factor, The previous heat correction factor. For heat renewal step size, This is the lateral deviation. Sensitivity to heat direction To prevent zero quantity, To correct the lower limit of calorie intake, This is the upper limit for heat correction. All the above parameters are dimensionless. Limits are applied based on the lower and upper limits of the heat correction: if the calculated result is lower than the lower limit, the lower limit is used; if it is higher than the upper limit, the upper limit is used; and if it is in between, the calculated result is retained.

[0095] Step S218 writes the updated moisture correction coefficient and heat correction coefficient back to the shared process model 118. The write-back operation overwrites the moisture correction coefficient and heat correction coefficient used in the next drying control cycle in the shared process model 118. The moisture reduction coefficient, temperature-induced drying coefficient, airflow-induced drying coefficient, temperature difference-induced heating coefficient, airflow-induced heating coefficient, and evaporative cooling coefficient remain at their calibration values ​​before the start of this batch. The material moisture content and material temperature at the current sampling time are used as the prediction starting point for the future prediction time domain, and the currently executed inlet air temperature and inlet air volume are used as the comparison benchmark for the continuity constraints of control actions.

[0096] Step S219 involves the control parameter output module 119 generating the future prediction time domain based on the shared process model 118 after writing back the moisture correction coefficient and the heat correction coefficient. For each control step covered by the prediction steps, the control parameter output module 119 substitutes the candidate values ​​of the inlet air temperature setpoint and the inlet air volume setpoint into the shared process model 118. The predicted moisture content and predicted material temperature obtained from the previous control step are used as the prediction starting point for the next control step, thus sequentially forming the predicted moisture content and predicted material temperature values ​​within the future prediction time domain. The candidate values ​​of the inlet air temperature setpoint and the inlet air volume setpoint are constrained by their corresponding upper and lower limits and the upper limit of a single change, respectively, while the predicted material temperature value is constrained by the upper limit of the material temperature.

[0097] The control parameter output module 119 uses the following objective function to evaluate candidate control sequences in the future prediction time domain:

[0098]

[0099] In the formula, , where is the dimensionless objective function value of the candidate control sequence; To predict the number of steps; This refers to the control step number within the future prediction time domain; This is the current sampling time; The moisture content weight is expressed as a percentage point negative square. For from the first Predicting the first sample time forward. Predicted moisture content for each control step; The target is the moisture content. Temperature action weights are measured in degrees Celsius negative square. This corresponds to the change in the inlet air temperature setpoint for the control step; The action weight of air volume is in the negative square of (cubic meters per hour); This corresponds to the change in the airflow setpoint for the control step. The first set of summations covers the first control step to the second. The second set of summations covers the control steps from the 0th to the 1st corresponding to the current output. One control step.

[0100]

[0101] In the formula, The lower limit of the inlet air temperature, The setpoint for the inlet air temperature corresponding to the control step within the future prediction time domain. This is the upper limit of the intake air temperature.

[0102]

[0103] In the formula, This is the lower limit of the air intake volume. The setpoint for the intake air volume corresponding to the control step within the future prediction time domain. This is the upper limit of the air intake volume.

[0104]

[0105] In the formula, This represents the predicted material temperature for the corresponding control step within the future prediction time domain. This represents the upper limit of the material temperature.

[0106]

[0107] In the formula, To correspond to the change in the inlet air temperature setpoint in the control step. This represents the upper limit of a single temperature change.

[0108]

[0109] In the formula, This corresponds to the change in the air intake setpoint for the control step. The upper limit of a single change in airflow is set. The control parameter output module 119 generates candidate control sequences step-by-step, starting from the actual airflow temperature and airflow setpoints of the previous drying control cycle, at discrete intervals of 1 degree Celsius for the inlet temperature setpoint and 50 cubic meters per hour for the inlet airflow setpoint. During generation, branches that do not meet the upper and lower limits of inlet temperature, upper and lower limits of inlet airflow, the upper limit of a single temperature change, the upper limit of a single airflow change, or the upper limit of material temperature are directly excluded. The objective function value of the remaining candidate control sequences is then calculated. The candidate control sequence with the smallest objective function value is selected. If multiple candidate control sequences have the same objective function value, the sequence with the smaller total change in adjacent control steps, the lower inlet temperature setpoint, and the lower inlet airflow setpoint is selected sequentially. After the solution is completed, only the inlet temperature and airflow setpoints corresponding to the first control step are output.

[0110] The measurement and control data of the pharmaceutical drying equipment 111 are all generated using 32-bit floating-point variables within the industrial controller. The drying process data acquisition module 112 reads the material moisture content, inlet air temperature, material temperature, and inlet air volume at the start of the cycle. The control parameter output module 119 writes the inlet air temperature setpoint to the setpoint variable in the inner loop of the inlet air heater actuator and the inlet air volume setpoint to the setpoint variable in the inner loop of the blower actuator before the end of the cycle, and reads the inlet air temperature and inlet air volume as actual execution inputs in the next cycle. All variables use the aforementioned engineering units and cycle numbers.

[0111] After acquiring the material moisture content and temperature in each drying control cycle, the industrial controller also performs batch endpoint determination. When the material moisture content is within the range determined by the target moisture content and the allowable difference in moisture content, and the material temperature does not exceed the upper limit of the material temperature, the industrial controller outputs a drug drying end control command. The corresponding determination relationship is as follows:

[0112]

[0113] In the formula, Given the current moisture content of the material, For the moisture content target, This represents the allowable difference in moisture content. This relationship is used to determine whether the current moisture content of the material is within the allowable range near the target moisture content.

[0114]

[0115] In the formula, The current material temperature, The upper limit of the material temperature. When the above two relationships are satisfied at the same time, the drug drying end control command causes the drug drying equipment 111 to switch to the predetermined end process; when the material moisture content is not within the range determined by the moisture content target and the moisture content allowable difference, or when the material temperature exceeds the upper limit of the material temperature, the industrial controller outputs the inlet air temperature setpoint and inlet air volume setpoint for the next drying control cycle, and continues to execute the drug drying control.

[0116] Example 1:

[0117] This embodiment uses a pharmaceutical drying device 111 for batch drying after wet granulation of solid oral dosage forms. The batch loading is 60 kg, the wet particle size is 0.3 mm to 1.0 mm, and the applicable material moisture content ranges from 2.0% to 15.0%. The drying control cycle is set to 30 seconds, and the prediction step count is set to 5. The material moisture content at the first sampling point of the batch is 11.00%, the inlet air temperature is 58.0 degrees Celsius, and the material temperature is 26.0 degrees Celsius. The initial moisture content thus preserved is 11.00%, and the initial thermal drive difference is 32.0 degrees Celsius. The target moisture content is set to 2.50%, the allowable moisture content difference is set to 0.20%, and the upper limit of the material temperature is set to 45.0 degrees Celsius. The lower limit of the inlet air temperature is set to 40.0 degrees Celsius, the upper limit of the inlet air temperature is set to 70.0 degrees Celsius, the lower limit of the inlet air volume is set to 700 cubic meters per hour, the upper limit of the inlet air volume is set to 1300 cubic meters per hour, the upper limit of a single temperature change is set to 2.0 degrees Celsius, and the upper limit of a single air volume change is set to 100 cubic meters per hour.

[0118] Six calibration batches, following the aforementioned parameter identification process, yielded the following coefficients: moisture reduction coefficient 0.020, temperature drying coefficient 0.010 percentage points per degree Celsius, airflow drying coefficient 0.0003 percentage points divided by cubic meters per hour, temperature difference heating coefficient 0.050, airflow heating coefficient 0.0005 degrees Celsius divided by cubic meters per hour, and evaporation cooling coefficient 0.40 degrees Celsius per percentage point. Validation data playback yielded initial values ​​of 1.00 for moisture and heat correction coefficients, 0.50 for lower and upper limits of moisture and heat correction, 1.50 for upper and upper limits of moisture and heat correction, a moisture update step size of 0.20, a heat update step size of 0.15, a zero-prevention amount of 0.001, and a minimum path movement of 0.005. The weighting for moisture content was set to 1.00 percentage points negative square, the weighting for temperature action was 0.50 degrees Celsius negative square, and the weighting for airflow action was 0.0005 cubic meters per hour negative square.

[0119] In the At the first sampling time, the drying process data acquisition module 112 obtained the following material moisture content: 6.20%, inlet air temperature: 60.0 degrees Celsius, material temperature: 34.0 degrees Celsius, and inlet air volume: 1000 cubic meters per hour. The moisture correction factor used at this sampling time was 1.05, and the heat correction factor was 0.95. The shared process model 118 obtained a moisture change baseline of 0.414 percentage points and the first... The predicted moisture content at the first sampling time was 5.7653%, and the baseline temperature change was 1.45 degrees Celsius. The predicted material temperature at the sampling time was 35.20362 degrees Celsius.

[0120] In the At the sampling time, the drying process data acquisition module 112 obtained a material moisture content of 5.85%, an inlet air temperature of 60.0 degrees Celsius, and a material temperature of 35.1 degrees Celsius. The actual drying movement direction formation module 113 obtained a thermal drive difference of 24.9 degrees Celsius, and thus obtained the first component of the measured path point (0.531818) and the second component (0.778125). The first component of the previous measured path point was 0.563636, and the second component was 0.812500. The first component of the path movement was -0.031818, and the second component was -0.034375. The Euclidean norm was approximately 0.04684, which is greater than the minimum path movement of 0.005. After normalization, the first component of the path forward vector was -0.679287, and the second component was -0.733873; the first component of the path lateral vector was 0.733873, and the second component was -0.679287.

[0121] The path prediction deviation generation module 114 obtains a predicted thermal drive difference of 24.79638 degrees Celsius based on the predicted values ​​of the current inlet air temperature and material temperature. Dividing the predicted moisture content by the initial moisture content, the first component of the predicted path point is 0.524118; dividing the predicted thermal drive difference by the initial thermal drive difference, the second component of the predicted path point is 0.774887. Subtracting the predicted path point from the measured path point yields a first component of the path prediction deviation of 0.007700 and a second component of 0.003238. This path prediction deviation is only sent to the path prediction deviation decomposition module 115.

[0122] The path prediction deviation decomposition module 115 obtains a forward deviation of -0.007607 and a lateral deviation of 0.003451, and writes them into the independent input ports of the moisture correction branch 116 and the heat correction branch 117, respectively. The two correction branches read the data from their respective independent input ports within the same drying control cycle.

[0123] Based on the current baseline moisture change, initial moisture content, initial thermal drive difference, and evaporation cooling coefficient, the first component of the moisture path sensitivity is -0.037636, and the second component is 0.005175. Projecting the moisture path sensitivity onto the path inverse vector yields a moisture directional sensitivity of 0.021768. Based on the current baseline temperature change and initial thermal drive difference, the first component of the heat path sensitivity is zero, and the second component is -0.045313. Projecting the heat path sensitivity onto the path transverse vector yields a heat directional sensitivity of 0.030780. The above calculations use the fixed model parameters and current state of the shared process model 118.

[0124] The moisture correction branch 116 substitutes the previous moisture correction coefficient of 1.05, forward deviation of -0.007607, moisture direction sensitivity of 0.021768, moisture update step size of 0.20, and zero-prevention amount of 0.001 into the moisture correction coefficient update relationship, obtaining an updated moisture correction coefficient of approximately 1.027530, which is between the lower and upper limits of moisture correction. The heat correction branch 117 substitutes the previous heat correction coefficient of 0.95, lateral deviation of 0.003451, heat direction sensitivity of 0.030780, heat update step size of 0.15, and zero-prevention amount of 0.001 into the heat correction coefficient update relationship, obtaining an updated heat correction coefficient of approximately 0.958182, which is between the lower and upper limits of heat correction.

[0125] The industrial controller writes the updated moisture correction factor of 1.027530 and heat correction factor of 0.958182 back to the shared process model 118, and uses the material moisture content of 5.85% and the material temperature of 35.1 degrees Celsius as the prediction starting point for the future prediction time domain. The control parameter output module 119 uses a moisture content weight of 1.00 percentage point negative square, a temperature action weight of 0.50 degrees Celsius negative square, and an air volume action weight of 0.0005 (cubic meters per hour) negative square to generate a candidate control sequence of five control steps at discrete intervals of an inlet air temperature setpoint of 1 degree Celsius and an inlet air volume setpoint of 50 cubic meters per hour.

[0126] Following the aforementioned layer-by-layer generation and constraint pruning process, the candidate control sequence with the minimum objective function value has an inlet air temperature setpoint of 61.0 degrees Celsius and an inlet air volume setpoint of 1050 cubic meters per hour for all five control steps. The predicted moisture content values ​​obtained by the shared process model 118 are 5.406107%, 4.971336%, 4.545501%, 4.128416%, and 3.719902%, respectively, and the predicted material temperature values ​​are 36.330971 degrees Celsius, 37.506616 degrees Celsius, 38.629510 degrees Celsius, 39.702109 degrees Celsius, and 40.726748 degrees Celsius, respectively, with an objective function value of 24.626934. The control parameter output module 119 writes the inlet air temperature setpoint of 61.0 degrees Celsius and the inlet air volume setpoint of 1050 cubic meters per hour corresponding to the first control step into the setpoint variable of the corresponding actuator inner loop.

[0127] When the material moisture content is 2.62% and the material temperature is 43.8 degrees Celsius in the subsequent drying control cycle, the material moisture content is within the range determined by the target moisture content of 2.50% and the allowable difference of moisture content of 0.20%, and the material temperature does not exceed the upper limit of material temperature of 45.0 degrees Celsius, the industrial controller outputs a control command to end the drying of the medicine.

[0128] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A process model based drying parameter adaptive control method, characterized in that, The method comprises: acquiring material moisture content, inlet air temperature and material temperature during drying, and forming an actual drying moving direction based on the material moisture content, the inlet air temperature and the material temperature; generating moisture content prediction values and material temperature prediction values based on a process model, and forming a path prediction deviation according to the moisture content prediction values, the material temperature prediction values, the material moisture content, the inlet air temperature and the material temperature; decomposing the path prediction deviation into a forward deviation and a lateral deviation according to the actual drying moving direction; setting a moisture correction coefficient and a heat correction coefficient, so that the forward deviation only enters a moisture correction branch to update the moisture correction coefficient, the lateral deviation only enters a heat correction branch to update the heat correction coefficient, and the path prediction deviation is not used as common update input of the moisture correction branch and the heat correction branch before decomposition; writing back the updated moisture correction coefficient and the heat correction coefficient to a shared process model; outputting an inlet air temperature setting value and an inlet air volume setting value executable by a pharmaceutical drying device based on the shared process model.

2. The process model based drying parameter adaptive control method according to claim 1, wherein, When forming the actual drying moving direction, an actual path point is formed according to the material moisture content and a heat driving difference determined by the difference between the inlet air temperature and the material temperature.

3. The process model based drying parameter adaptive control method according to claim 2, wherein, When forming the actual path point, an initial moisture content and an initial heat driving difference corresponding to a first sampling point are saved as dimensionless reference, and the material moisture content and the heat driving difference of subsequent sampling points are processed by dimensionless treatment based on the initial moisture content and the initial heat driving difference.

4. The process model based drying parameter adaptive control method of claim 2, wherein, The actual drying moving direction comprises a path forward vector and a path lateral vector, the path forward vector is determined according to the path movement amount of the actual path point at adjacent sampling time, and the path lateral vector is determined according to the path forward vector and is perpendicular to the path forward vector.

5. The process model based drying parameter adaptive control method according to claim 2, wherein, When forming the path prediction deviation, a predicted heat driving difference is formed according to the inlet air temperature and the material temperature prediction values, and a predicted path point is formed from the moisture content prediction values and the predicted heat driving difference; the path prediction deviation is formed according to the difference between the actual path point and the predicted path point.

6. The process model based drying parameter adaptive control method according to claim 4, wherein, The moisture correction branch and the heat correction branch are respectively provided with independent input ports, and only the independent input port corresponding to the forward deviation is called to update the moisture correction coefficient, and only the independent input port corresponding to the lateral deviation is called to update the heat correction coefficient when performing model correction.

7. The process model based drying parameter adaptive control method according to claim 6, wherein, When updating the moisture correction coefficient and the heat correction coefficient, a moisture direction sensitivity represents the change relationship of the predicted path point along the path forward vector when the moisture correction coefficient changes, a heat direction sensitivity represents the change relationship of the predicted path point along the path lateral vector when the heat correction coefficient changes, the moisture correction branch updates the moisture correction coefficient based on the forward deviation and the moisture direction sensitivity, and the heat correction branch updates the heat correction coefficient based on the lateral deviation and the heat direction sensitivity.

8. The process model based drying parameter adaptive control method of claim 1, wherein, When outputting the inlet air temperature setting value and the inlet air volume setting value, a future prediction time domain is generated based on the shared process model after writing back the moisture correction coefficient and the heat correction coefficient, and the inlet air temperature setting value and the inlet air volume setting value corresponding to a first control step are selected from the future prediction time domain.

9. The process model based drying parameter adaptive control method of claim 1, wherein, When the material moisture content is within a range determined by a moisture content target and a moisture content allowable deviation, and the material temperature does not exceed a material temperature upper limit, a pharmaceutical drying end control instruction is output. When the moisture content of the material is not in the range determined by the moisture content target and the moisture content allowable difference or the material temperature exceeds the material temperature upper limit, the supply air temperature setting value and the supply air amount setting value for the next drying control period are output.