An artificial intelligence-based hippurate urotropin effervescent tablet prescription and preparation parameter joint optimization method

Abstract

This invention discloses an artificial intelligence-based method for joint optimization of prescription and preparation parameters of urotropin hippuric acid effervescent tablets, belonging to the field of prescription and preparation parameter optimization technology. It includes: classifying data based on urotropin hippuric acid effervescent tablets to obtain at least two data categories, resulting in summary information items; integrating data based on the summary information items, separately integrating prescription variables, process parameter variables, and quality attribute variables to obtain prescription variable sets, process parameter variable sets, and quality attribute variable sets. This invention sets interactive tests of change magnitude values ​​to quantify the impact of prescription variables and process parameters on quality attributes, ensuring that optimization focuses on core variables and avoids interference from secondary factors. Furthermore, it constructs prescription-process interaction modeling and process-quality transfer modeling. The models, supplemented by data-driven models and cost models, achieve multi-objective simultaneous optimization.

Description

Technical Field

[0001] This invention relates to the field of prescription and preparation parameter optimization technology, specifically to an artificial intelligence-based method for joint optimization of prescription and preparation parameters of hippuric acid urotropin effervescent tablets. Background Technology

[0002] Urotropin hippurate effervescent tablets are a type of medicine commonly used to promote diuresis and relieve urinary tract-related discomfort. The main component is hippuric acid, also known as sodium hippurate, which is usually a component of the drug and is used to regulate the body's acid-base balance or in combination with other ingredients. In the preparation of urotropin hippurate effervescent tablets, the formulation or preparation parameters need to be optimized to improve efficacy and bioavailability. At the same dosage, disintegration and dissolution rates affect the rate and extent of drug absorption, while also improving stability and shelf life. Temperature, humidity, mixing uniformity, and compression parameters all affect the tablet's disintegration time, water content, hardness, and disintegration properties, thus impacting efficacy and safety.

[0003] A machine learning-based prescription optimization method, patent publication number CN118866262A, assigns weights to the Chinese herbs in the prescription based on syndrome elements. This digitizes the theoretical experience of renowned veteran TCM experts in symptom elements, guiding the reduction of prescription herbs. It can provide guidance for prescription optimization for ordinary physicians and serve as a carrier for the dissemination and development of the experience of renowned veteran TCM experts, thus realizing the preservation and inheritance of their experience.

[0004] In the process of optimizing formulations or preparation parameters, the above-mentioned and similar technical solutions often involve trade-offs between dissolution, disintegration and stability in traditional effervescent tablets. Furthermore, the preparation process is sensitive to factors such as temperature, humidity, tableting force and excipient particle size, making it difficult to predict quality fluctuations and scale-up failure modes. Moreover, when optimizing formulations or preparation parameters, they are often single-objective optimizations or based on empirical rules, making it difficult to simultaneously consider the coupling effect of formulation and process parameters as well as cost constraints. As a result, the formulation optimization and preparation parameter optimization processes cannot be combined synchronously, which easily leads to stability prediction failures. Summary of the Invention

[0005] The purpose of this invention is to provide an artificial intelligence-based method for jointly optimizing the formulation and preparation parameters of hippuric acid urotropin effervescent tablets, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for jointly optimizing the formulation and preparation parameters of hippuric acid urotropin effervescent tablets based on artificial intelligence, comprising:

[0007] Based on the data classification of hippuric urotropin effervescent tablets, at least two data categories were obtained, and summary information items were obtained.

[0008] Data is integrated based on the summary information items, and prescription variables, process parameter variables and quality attribute variables are integrated respectively to obtain prescription variable set, process parameter variable set and quality attribute variable set;

[0009] Based on reinforcement learning, data association is performed on the prescription variable set, process parameter variable set, and quality attribute variable set to obtain key data mapping and key prescription items, key process parameter items, and key quality attribute items.

[0010] A deep learning hybrid architecture is adopted to construct a coupled model based on key prescription items, key process parameter items, and key quality attribute items to obtain a modeling information set.

[0011] Coupled core equations are created based on the modeling information set, and the quality attribute output is obtained by creating a model integration using the interaction matrix.

[0012] Model validation and parameter calibration are performed based on the output data items, thereby enabling coordinated adjustment of key formulation items, key process parameters, and key quality attributes, achieving joint optimization of formulation data and preparation parameter data.

[0013] Furthermore, the method for obtaining the summary information items includes:

[0014] Obtain preparation information for hippuric acid urotropin effervescent tablets, including pre-preparation data, in-preparation data, and post-preparation data.

[0015] The data information before preparation includes the type of material, the data information during preparation includes the preparation parameters, and the data information after preparation includes the stability of the finished product. The data information before preparation, the data information during preparation, and the data information after preparation are combined to form a summary information item.

[0016] Furthermore, the methods for obtaining the prescription variable set, process parameter variable set, and quality attribute variable set include:

[0017] Prescription variables include the mass ratio of active pharmaceutical ingredient to excipients, type and amount of disintegrant, and proportion of swelling agent, to obtain the first information set. Based on the first information set, information is integrated and data is classified to obtain the prescription variable set.

[0018] The process parameter variables include mixing time, tableting force, drying temperature, and effervescent initiation humidity control, resulting in a second information set. Based on the second information set, information is integrated and data is classified to obtain a process parameter variable set.

[0019] The quality attribute variables include the target value of dissolution curve similarity factor, the target value of disintegration time, the target value of particle uniformity, and the target value of unit production cost, which yields a third information set. Based on the third information set, information is integrated and data is classified to obtain the quality attribute variable set.

[0020] Furthermore, the method for obtaining the prescription variable set, process parameter variable set, and quality attribute variable set also includes: setting participation thresholds based on the first information set, the second information set, and the third information set respectively, where the participation threshold is the percentage value of variable participation; filtering the first information set, the second information set, and the third information set based on the participation threshold, removing variables that do not reach the participation threshold, and obtaining the first filtered information set, the second filtered information set, and the third filtered information set, thereby obtaining the prescription variable set, the process parameter variable set, and the quality attribute variable set.

[0021] Furthermore, the method for obtaining the key prescription items, key process parameter items, and key quality attribute items includes:

[0022] Set a change range value, which is a fixed percentage value. Using the process parameter variable set as a fixed attribute, interact sequentially with the independent variables in the prescription variable set based on the change range value to obtain the number and value of changes in the independent variables in the quality attribute variable set, and obtain the first change information set.

[0023] Using the prescription variable set as a fixed attribute, the independent variables in the process parameter variable set are sequentially interacted with based on the change range value to obtain the number and value of changes of the independent variables in the quality attribute variable set, thus obtaining the second change information set;

[0024] Based on the first change information set and the second change information set, the independent variable with the highest number of changes in the quality attribute variable set is obtained to obtain the control variable item. The independent variables corresponding to the control variable item in the first change information set and the second change information set are used as key process parameter items and key prescription items.

[0025] Based on the first and second change information sets, the independent variables with the highest number of changes in the quality attribute variable set are obtained to obtain the demand variable items, which are then used as the key quality attribute items.

[0026] Furthermore, the method for obtaining the modeling information set includes:

[0027] The coupled model construction includes formulation and process interaction modeling as well as process and quality transfer modeling. The mechanistic equation for formulation and process interaction modeling is to quantify the physical relationship between tableting force and disintegration time, while the kinetic equation for process and quality transfer modeling is the coupling effect of drying temperature and humidity on dissolution.

[0028] Furthermore, the coupled model construction also includes data-driven supplementary modeling and cost model construction. The methods for obtaining the modeling information set include: data-driven supplementary modeling for predicting particle uniformity using a random forest model, and cost model construction for calculating unit production cost using linear regression.

[0029] Furthermore, the method for obtaining the output data item includes:

[0030] The key interactions to obtain the interaction matrix include the interaction between tableting force and disintegrant type and disintegration time, and the interaction between drying and swelling agent ratio and particle uniformity.

[0031] Furthermore, the method for model validation and parameter calibration includes:

[0032] Set the data volume and test metrics, including quality attributes and test range. Also set the experimental matrix, the number of times the center point is repeated, and obtain the calibration results.

[0033] Compared with the prior art, the beneficial effects of the present invention are:

[0034] This AI-based method for jointly optimizing the formulation and preparation parameters of urotropin hippuric acid effervescent tablets identifies key formulation items and key quality attributes by classifying data types, such as material types in pre-preparation data, temperature and humidity in preparation data, disintegration time in post-preparation data, and using reinforcement learning analysis. It sets interactive tests for change magnitude values, quantifies the impact of formulation variables and process parameters on quality attributes, ensures that optimization focuses on core variables and avoids interference from secondary factors, and constructs interactive modeling of formulation and process and transfer modeling of process and quality. The model, supplemented by data-driven and cost models, achieves simultaneous optimization of multiple objectives. Attached Figure Description

[0035] Figure 1 This is a flowchart illustrating the overall execution process of the present invention.

[0036] Figure 2 This is a schematic diagram of the key data mapping process of the present invention;

[0037] Figure 3 This is a schematic diagram of the model construction and verification process of the present invention;

[0038] Figure 4 This is a schematic diagram of the key variable screening and optimization process of the present invention;

[0039] Figure 5 This is a schematic diagram of the variable interaction and key item identification process of the present invention. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Traditional effervescent tablets inherently involve a trade-off between dissolution, disintegration, and stability. To ensure rapid drug release in the body, effervescent tablets typically require excellent disintegration properties, often achieved by adding large amounts of disintegrants or using readily soluble excipients. However, these measures may reduce the stability of effervescent tablets, making them more susceptible to deliquescence, chemical reactions, or physical deterioration. Conversely, improving the stability of effervescent tablets may sacrifice disintegration properties, leading to slower dissolution rates and affecting drug bioavailability. This trade-off complicates formulation optimization, requiring the search for a balance among multiple objectives. Single-objective optimization often fails to meet practical needs. Furthermore, the preparation process of effervescent tablets is highly sensitive to various factors such as temperature, humidity, compression force, and excipient particle size. For example, excessively high temperature and humidity may cause excipients to become damp, affecting tablet formation and subsequent disintegration properties, while excessive compression force... This can lead to effervescent tablets being too hard and disintegrating slowly; insufficient compression force can result in loose, brittle tablets; and uneven particle size of excipients can cause uneven mixing, affecting the quality uniformity of the effervescent tablets. These factors interact with each other, making the preparation process difficult to control and quality fluctuations unpredictable. The technical solution provided in this application, by classifying data types (e.g., material type in pre-preparation data, temperature and humidity in preparation data, disintegration time in post-preparation data) and using reinforcement learning analysis, identifies key formulation items and key quality attributes. It sets change magnitude values ​​for interactive testing, quantifies the impact of formulation variables and process parameters on quality attributes, ensures optimization focuses on core variables, avoids interference from secondary factors, and constructs interaction modeling between formulation and process and transfer modeling between process and quality. The model, through data-driven supplementation and cost modeling, achieves multi-objective simultaneous optimization, such as... Figure 1 As shown, it includes steps S100-S600.

[0042] Step S100: Classify data types based on hippuric acid urotropine effervescent tablets.

[0043] It is important to note that data classification should yield at least two data categories to obtain summary information items. The methods for obtaining summary information items include: obtaining preparation information for hippuric acid urotropin effervescent tablets, including pre-preparation data, in-preparation data, and post-preparation data; pre-preparation data includes material types, in-preparation data includes preparation parameters, and post-preparation data includes finished product stability. The summary information items are obtained by combining the pre-preparation data, in-preparation data, and post-preparation data.

[0044] Specifically, during the preparation of hippuric acid urotropine effervescent tablets, three types of data were obtained: pre-preparation data, in-preparation data, and post-preparation data. Pre-preparation data includes information such as the type, quantity, and ratio of raw materials. In-preparation data includes preparation parameter information, such as preparation temperature and humidity, and preparation pressure. Post-preparation data includes finished product stability information, such as the disintegration time and particle uniformity of the effervescent tablets. Therefore, the pre-preparation, in-preparation, and post-preparation data are summarized to obtain a summary information item.

[0045] Step S200: Integrate the data based on the summary information items, and integrate the prescription variables, process parameter variables and quality attribute variables respectively.

[0046] It is important to note that the integration of prescription variables, process parameter variables, and quality attribute variables yields the prescription variable set, process parameter variable set, and quality attribute variable set, respectively. The acquisition methods include: prescription variables, such as the mass ratio of active pharmaceutical ingredient to excipients, disintegrant type and dosage, and bulking agent ratio, yielding a first information set. Information integration and data classification are then performed based on this first information set to obtain the prescription variable set. Process parameter variables, including mixing time, tableting force, drying temperature, and effervescent initiation humidity control, yield a second information set. Information integration and data classification are then performed based on this second information set to obtain the process parameter variable set. Quality attribute variables, including target values ​​for dissolution curve similarity factor, disintegration time, particle uniformity, and unit production cost, yield a third information set. Information integration and data classification are then performed based on this third information set to obtain the quality attribute variable set.

[0047] Specifically, after obtaining the summarized information items, the information is integrated, including prescription variable information, process parameter variable information, and quality attribute variable information. Prescription variables include the mass ratio of active pharmaceutical ingredient to excipient, disintegrant type and dosage, and swelling agent ratio. Process parameter variables include mixing time, tableting force, drying temperature, and effervescent initiation humidity control. Quality attribute variables include target values ​​for dissolution curve similarity factor, disintegration time, particle uniformity, and unit production cost. This results in the prescription variable set, process parameter variable set, and quality attribute variable set.

[0048] It should be noted that the method for obtaining the prescription variable set, process parameter variable set, and quality attribute variable set also includes: setting participation thresholds based on the first information set, the second information set, and the third information set respectively. The participation threshold is the percentage value of the variable participation. Based on the participation threshold, the first information set, the second information set, and the third information set are filtered to remove variables that do not reach the participation threshold, thereby obtaining the first filtered information set, the second filtered information set, and the third filtered information set, and then obtaining the prescription variable set, the process parameter variable set, and the quality attribute variable set.

[0049] Specifically, such as Figure 4 As shown, in the preparation of hippuric acid urotropine effervescent tablets, different participation values ​​exist in the formulation variables, such as the mass ratio of active pharmaceutical ingredient to excipients, the type and amount of disintegrant, and the proportion of swelling agent. Different participation values ​​also exist in the process parameter variables, such as mixing time, tableting force, drying temperature, and effervescent initiation humidity control. Similarly, different participation values ​​exist in the quality attribute variables, such as the target value of dissolution curve similarity factor, disintegration time, particle uniformity, and unit production cost. For example, if the active pharmaceutical ingredient accounts for 50% of the formulation variables, the mass ratio of excipients is 20%, the amount of disintegrant and the proportion of swelling agent each account for 10%, secondary excipient a accounts for 9%, and secondary excipient b accounts for 1%, then by setting a participation threshold of 5%, data screening is performed based on this threshold. Variables that do not reach the participation threshold are eliminated, i.e., secondary excipient b is removed, thus obtaining the first screening information set. This first screening information set is used as the formulation variable set, thereby achieving the retention of important variables and the elimination of non-important variables, improving the accuracy of the optimization focus of the formulation and preparation parameters.

[0050] Step S300: Using reinforcement learning, perform data association on the prescription variable set, process parameter variable set, and quality attribute variable set to obtain key data mapping.

[0051] It is important to note that, such as Figures 2-3As shown, key data mapping yields key prescription items, key process parameter items, and key quality attribute items. The acquisition method includes: setting a change range value, where the change range value is a fixed percentage value; using the process parameter variable set as a fixed attribute, sequentially interacting with the independent variables in the prescription variable set based on the change range value to obtain the number and value of changes in the independent variables in the quality attribute variable set, thus obtaining a first change information set; using the prescription variable set as a fixed attribute, sequentially interacting with the independent variables in the process parameter variable set based on the change range value to obtain the number and value of changes in the independent variables in the quality attribute variable set, thus obtaining a second change information set; based on the first and second change information sets, identifying the independent variable with the highest number of changes in the quality attribute variable set to obtain a control variable item; and using the independent variables corresponding to the control variable item in the first and second change information sets as key process parameter items and key prescription items; and based on the first and second change information sets, identifying the independent variable with the highest number of changes in the quality attribute variable set to obtain a demand variable item, which is then used as a key quality attribute item.

[0052] Specifically, such as Figure 5As shown, the set change range is 1%. First, the process parameter variable set is kept constant, i.e., the mass ratio of active pharmaceutical ingredient to excipients, the type and amount of disintegrant, and the proportion of swelling agent. Based on the change range value, it interacts sequentially with the independent variables in the prescription variable set, i.e., adjusting the mixing time, compression force, drying temperature, and effervescent initiation humidity control sequentially by the change range. The number and value of changes in the independent variables in the quality attribute variable set are obtained, resulting in the first change information set. Then, the prescription variable set is kept constant again, i.e., the mixing time, compression force, drying temperature, and effervescent initiation humidity control are kept constant. Based on the change range value, it interacts sequentially with the independent variables in the process parameter variable set, i.e., adjusting the mass ratio of active pharmaceutical ingredient to excipients, the type and amount of disintegrant, and the proportion of swelling agent sequentially by the change range. The number and value of changes in the independent variables in the quality attribute variable set are obtained, resulting in the second change information set. Information sets are used to identify the independent variables with the highest number of changes in the quality attribute variable set, based on the first and second change information sets. When the mixing time, tableting force, drying temperature, effervescent initiation humidity control, and the mass ratio of active pharmaceutical ingredient to excipients, disintegrant type and dosage, and swelling agent ratio are adjusted sequentially by the change magnitude, the target values ​​of dissolution curve similarity factor, disintegration time, particle uniformity, and unit production cost in the quality attribute variable set will all change. At this time, the independent variable with the highest number of changes is selected as the control variable. The independent variables corresponding to the control variable in the first and second change information sets are then selected as key process parameters and key formulation items. At the same time, based on the first and second change information sets, the independent variable with the highest number of changes in the quality attribute variable set is selected to obtain the demand variable, which is used as the key quality attribute item.

[0053] In the specific implementation process, it is now necessary to produce hippuric acid urotropine effervescent tablets. Among the set formulation variables, the mass ratio of active pharmaceutical ingredient to excipients is 0.4, the disintegrant type is sodium carboxymethyl starch at a dosage of 6%, and the swelling agent ratio is 10%. Among the process parameter variables, the mixing time is 22 min, the tableting force is 12 kN, the drying temperature is 60℃, and the effervescent initiation humidity is controlled at 45%. Among the quality attribute variables, the target value for the dissolution curve similarity factor is 54, the target value for disintegration time is 98 s, the target value for particle uniformity is 4.2%, and the target value for unit production cost is 1.5 yuan. At this time, the first change information set and the second change information set are obtained according to the set change range, as shown in Table 1 and Table 2 respectively:

[0054] Table 1

[0055]

[0056] As shown in Table 1, the independent variable with the highest number of changes in the set of quality attribute variables corresponds to humidity control in the first set of change information, thus yielding the key process parameter item.

[0057] Table 2

[0058]

[0059] As shown in Table 2, the independent variable with the highest number of changes in the set of quality attribute variables corresponds to the independent variable in the second set of change information as the mass ratio of active pharmaceutical ingredient to excipients, thus yielding the key prescription item;

[0060] In the first and second change information sets, the independent variable with the highest number of changes in the quality attribute variable set is the disintegration time target value, which is then used as the demand variable item, with the disintegration time target value as the key quality attribute item.

[0061] Step S400: A deep learning hybrid architecture is used to construct a coupled model based on key prescription items, key process parameter items, and key quality attribute items to obtain a modeling information set.

[0062] It should be noted that the methods for obtaining the modeling information set include: the construction of coupled models includes the interaction modeling of formulation and process and the modeling of process and quality transfer. The mechanistic equation for the interaction modeling of formulation and process is to quantify the physical relationship between tableting force and disintegration time, and the kinetic equation for the modeling of process and quality transfer is the coupling effect of drying temperature and humidity on dissolution.

[0063] Specifically, the mechanistic equation for the interaction modeling of formulation and process quantifies the physical relationship between compression force and disintegration time, which is manifested as follows:

[0064] ;

[0065] in For disintegration time, The coefficient representing the influence of tableting force is 8.2, which is the experimental calibration value. For tablet compression force, The porosity decay index, fitted by CT scan, is 0.75. For negative correction of disintegrant dosage, when the disintegrant is sodium carboxymethyl starch. It is −0.8. This refers to the dosage of disintegrant. The positive correction for the bentonite ratio is 0.3. This refers to the proportion of the swelling agent. Let N be the residual term, which follows a normal distribution N(0, 1.2). 2 );

[0066] The kinetic equation for process and quality transfer modeling is the coupling effect of drying temperature and humidity on dissolution, specifically manifested as follows:

[0067] ;

[0068] in For dissolution similarity factor, The maximum dissolution factor is 72.5. The activation energy for the hydrolysis of hippuric acid is 85.6 kJ / mol. The ideal gas constant is 0.008314 kJ / mol·K. The drying temperature, The humidity sensitivity coefficient, fitted using accelerated testing, is set to 0.015. Humidity.

[0069] It is important to note that the construction of the coupled model also includes data-driven supplementary modeling and cost model construction. The methods for obtaining the modeling information set include: data-driven supplementary modeling for predicting particle uniformity using the random forest model, and cost model construction for calculating unit production cost using linear regression.

[0070] Specifically, the uniformity of particles is predicted using a random forest model, with the number of hyperparameter trees n set. trees =200, maximum depth d max =8, feature sampling ratio f ratio =0.7, input mixing time, disintegrant dosage, and bulking agent ratio, output particle uniformity, and construct a cost model to calculate unit production cost using linear regression. The formula is:

[0071] ;

[0072] in Unit production cost The base cost is 0.12. The cost of cross-linked carboxymethyl cellulose is set to 0.1. It is a disintegrant, specifically cross-linked carboxymethyl cellulose. This is the cost coefficient for disintegrant dosage, with a value of 0.002. This refers to the dosage of disintegrant. The mixed time cost coefficient has a value of 0.0005. For mixed time.

[0073] Step S500: Create coupled core equations based on the modeling information set, and use the interaction matrix to create model integration to output quality attributes, thereby obtaining output data items.

[0074] It is important to note that the methods for obtaining the output data items include: obtaining the key interactions of the interaction matrix, including the interaction between tableting force and disintegrant type and disintegration time, as well as the interaction between drying temperature and swelling agent ratio and particle uniformity.

[0075] Specifically, the interaction matrix includes the interaction between tableting force and disintegrant type on disintegration time and the interaction between drying temperature and swelling agent ratio on particle uniformity, with value ranges of [-1.2, 0.8] and [0.05, 0.15], respectively.

[0076] Step S600: Perform model validation and parameter calibration based on the output data items.

[0077] It is important to note that by coordinating the adjustment of key prescription items, key process parameters, and key quality attributes, joint optimization of prescription data and preparation parameter data was achieved. The methods for model validation and parameter calibration include: setting the data volume and test indicators, including quality attributes and test ranges, setting the experimental matrix, setting the number of repetitions of the center point, and obtaining calibration results.

[0078] Specifically, the minimum data requirement is 50 independent experiments, covering the variable range boundaries. The test indicators include target values ​​for dissolution curve similarity factor, disintegration time, and particle uniformity. The target values ​​for dissolution curve similarity factor are ≤3.0, disintegration time is ≤4s, and particle uniformity is ≥0.9. A Box-Behnken design is used with 6 repetitions at the center point and 5 factors, namely the mass ratio of active pharmaceutical ingredient to excipients, disintegrant dosage, mixing time, tableting force, and drying temperature.

[0079] In the specific implementation process, the experimental matrix is ​​set as shown in Table 3:

[0080] Table 3

[0081]

[0082] The parameter coding rules are as follows: tableting force: low level 8kN (-1), baseline 12kN (0), high level 16kN (+1); drying temperature: 50℃ (-1), 60℃ (0), 70℃ (+1), at which point the key coefficients in the disintegration time model are calibrated:

[0083]

[0084] in For disintegration time, The coefficient representing the influence of tableting force is 8.2, which is the experimental calibration value. For tablet compression force, For negative correction of disintegrant dosage, when the disintegrant is sodium carboxymethyl starch. It is −0.8. The disintegrant dosage was determined by repeating the experiment at the center point, and the results are shown in Table 4.

[0085] Table 4

[0086]

[0087] The baseline value for the measured disintegration time was calculated to be 93.3 s. Coefficients were fitted using the least squares method. Experiment 3 data: tableting force 16 kN, disintegrant dosage 6%, disintegration time 113.5 s; Experiment 7 data: tableting force 7 kN, disintegrant dosage 8%, disintegration time 85.2 s. The fitted equation is:

[0088] ;

[0089] Solving for α, we get α = 8.15, which is 0.6% lower than the initial value of 8.2; and γ = −0.82, which is 2.5% lower than the initial value of −0.8.

[0090] Then, prediction accuracy tests are conducted, and key prescription items, key process parameters, and key quality attributes are adjusted in a coordinated manner based on the test results.

[0091] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended embodiments and their equivalents.

Claims

1. An artificial intelligence-based method for joint optimization of the formulation and preparation parameters of hippuric acid urotropin effervescent tablets, comprising: Based on the data classification of hippuric urotropin effervescent tablets, at least two data categories were obtained, and summary information items were obtained. Data is integrated based on the summary information items, and prescription variables, process parameter variables and quality attribute variables are integrated respectively to obtain prescription variable set, process parameter variable set and quality attribute variable set; Its features are: Reinforcement learning is used to perform data association on the prescription variable set, process parameter variable set, and quality attribute variable set to obtain key data mapping and obtain key prescription items, key process parameter items, and key quality attribute items. A deep learning hybrid architecture is adopted to construct a coupled model based on key prescription items, key process parameter items, and key quality attribute items to obtain a modeling information set. Coupled core equations are created based on the modeling information set, and the quality attribute output is obtained by creating a model integration using the interaction matrix. Model validation and parameter calibration are performed based on the output data items, thereby enabling coordinated adjustment of key formulation items, key process parameters, and key quality attributes, achieving joint optimization of formulation data and preparation parameter data.

2. The method for the prescription and preparation parameter optimization of the artificial intelligence-based methenamine hippurate effervescent tablets according to claim 1, characterized in that: The methods for obtaining the summary information items include: Obtain preparation information for hippuric acid urotropin effervescent tablets, including pre-preparation data, in-preparation data, and post-preparation data. The data information before preparation includes the type of material, the data information during preparation includes the preparation parameters, and the data information after preparation includes the stability of the finished product. The data information before preparation, the data information during preparation, and the data information after preparation are combined to form a summary information item.

3. The method for the prescription and preparation parameter optimization of the artificial intelligence-based methenamine hippurate effervescent tablets according to claim 1, characterized in that: The methods for obtaining the prescription variable set, process parameter variable set, and quality attribute variable set include: Prescription variables include the mass ratio of active pharmaceutical ingredient to excipients, type and amount of disintegrant, and proportion of swelling agent, to obtain the first information set. Based on the first information set, information is integrated and data is classified to obtain the prescription variable set. The process parameter variables include mixing time, tableting force, drying temperature, and effervescent initiation humidity control, resulting in a second information set. Based on the second information set, information is integrated and data is classified to obtain the process parameter variable set. The quality attribute variables include the target value of dissolution curve similarity factor, the target value of disintegration time, the target value of particle uniformity, and the target value of unit production cost, which yields a third information set. Based on the third information set, information is integrated and data is classified to obtain the quality attribute variable set.

4. The method for jointly optimizing the formulation and preparation parameters of hippuric acid urotropin effervescent tablets based on artificial intelligence according to claim 3, characterized in that: The method for obtaining the prescription variable set, process parameter variable set, and quality attribute variable set also includes: setting participation thresholds based on the first information set, the second information set, and the third information set respectively, wherein the participation threshold is the percentage value of variable participation; filtering the data of the first information set, the second information set, and the third information set based on the participation threshold, and removing variables that do not reach the participation threshold to obtain the first filtered information set, the second filtered information set, and the third filtered information set, thereby obtaining the prescription variable set, the process parameter variable set, and the quality attribute variable set.

5. The method for jointly optimizing the formulation and preparation parameters of hippuric acid urotropin effervescent tablets based on artificial intelligence according to claim 1, characterized in that: The methods for obtaining the key prescription items, key process parameter items, and key quality attribute items include: Set a change range value, which is a fixed percentage value. Using the process parameter variable set as a fixed attribute, interact sequentially with the independent variables in the prescription variable set based on the change range value to obtain the number and value of changes in the independent variables in the quality attribute variable set, and obtain the first change information set. Using the prescription variable set as a fixed attribute, the independent variables in the process parameter variable set are sequentially interacted with based on the change range value to obtain the number and value of changes of the independent variables in the quality attribute variable set, thus obtaining the second change information set; Based on the first change information set and the second change information set, the independent variable with the highest number of changes in the quality attribute variable set is obtained to obtain the control variable item. The independent variables corresponding to the control variable item in the first change information set and the second change information set are used as key process parameter items and key prescription items. Based on the first and second change information sets, the independent variables with the highest number of changes in the quality attribute variable set are obtained to obtain the demand variable items, which are then used as the key quality attribute items.

6. The method for jointly optimizing the formulation and preparation parameters of hippuric acid urotropin effervescent tablets based on artificial intelligence according to claim 1, characterized in that: The method for obtaining the modeling information set includes: The coupled model construction includes formulation and process interaction modeling as well as process and quality transfer modeling. The mechanistic equation for formulation and process interaction modeling is to quantify the physical relationship between tableting force and disintegration time, while the kinetic equation for process and quality transfer modeling is the coupling effect of drying temperature and humidity on dissolution.

7. The method for jointly optimizing the formulation and preparation parameters of hippuric acid urotropin effervescent tablets based on artificial intelligence according to claim 6, characterized in that: The coupled model construction also includes data-driven supplementary modeling and cost model construction. The methods for obtaining the modeling information set include: data-driven supplementary modeling for predicting particle uniformity using a random forest model, and cost model construction for calculating unit production cost using linear regression.

8. The method for jointly optimizing the formulation and preparation parameters of hippuric acid urotropin effervescent tablets based on artificial intelligence according to claim 1, characterized in that: The method for obtaining the output data items includes: The key interactions to obtain the interaction matrix include the interaction between tableting force and disintegrant type and disintegration time, and the interaction between drying and swelling agent ratio and particle uniformity.

9. The method for jointly optimizing the formulation and preparation parameters of hippuric acid urotropin effervescent tablets based on artificial intelligence according to claim 1, characterized in that: The methods for model validation and parameter calibration include: Set the data volume and test metrics, including quality attributes and test range. Also set the experimental matrix, the number of times the center point is repeated, and obtain the calibration results.

Images