Thermoplastic coated controlled release tablet preparation process parameter setting system and method

Abstract

The invention relates to the technical field of tablet preparation, in particular to a thermoplastic coated controlled release tablet preparation process parameter setting system and method.The method comprises the steps that parameter information is collected through a data processing module, and invalid samples are removed to expand a data set; the model building module takes a support vector machine as a core classifier and builds a parameter space containing an RBF kernel and a linear kernel; the model training module is used for processing continuous parameters through standardization and generating a high-accuracy prediction model in combination with grid search and three-fold cross validation; and the parameter setting module generates a candidate parameter combination based on the prediction result and visually outputs an optimal scheme. The method comprises the steps of data processing, model construction, training and parameter setting, solves the problems of quality fluctuation, small sample overfitting, difficulty in description of a nonlinear relationship and the like caused by dependence on artificial experience in a traditional heat sealing process, realizes intelligent optimization of production parameters of the thermoplastic coated controlled release tablets, and remarkably improves the production stability and efficiency.

Description

Technical Field

[0001] The present invention belongs to the technical field of tablet preparation, and in particular relates to a system and method for setting process parameters for the preparation of thermoplastic-coated controlled-release tablets. Background Art

[0002] In the preparation of osmotic pump controlled-release tablets, traditional organic solvent coating processes pose challenges such as flammability and explosion risks, environmental pollution, and high costs. Experimental results demonstrate that the release characteristics of osmotic pump controlled-release tablets prepared using a novel, zero-organic solvent thermoplastic coating method are consistent with those of currently available organic solvent spray-coated osmotic pump controlled-release tablets. The production process for this novel thermoplastic coating method includes unwinding the lower coating film, heating, bubble forming, tablet filling, covering with the upper coating film, heat sealing the film, cutting, and waste roll recycling. The heat sealing process is the core technology, and its quality directly impacts the product performance and consistent release rate of the thermoplastic-coated controlled-release tablets.

[0003] The existing technology has the following problems: (1) The processing parameters of the heat sealing process, such as temperature, plasticizer content, and time, are set based on the operator's experience and are significantly affected by subjective factors, which may lead to fluctuations in the quality of thermoplastic coated controlled-release tablets in each production batch. (2) Traditional methods lack a prediction mechanism for process parameters and heat sealing effects, and are unable to evaluate the impact of parameter settings on the quality of finished products in advance. Parameters can only be adjusted through trial and error, which easily leads to waste of raw materials and low production efficiency. (3) The data collection cost of tablet coating production lines in the pharmaceutical field is high, resulting in a relatively small amount of sample data for the prediction of thermoplastic coated controlled-release tablet preparation parameters and insufficient generalization, forming a small sample scenario. In this scenario, machine learning models are prone to overfitting, and the heat sealing effect and process parameters show a strong nonlinear and coupled relationship. Traditional linear modeling methods such as partial least squares (PLS) cannot accurately describe this complex relationship, resulting in insufficient model prediction accuracy.

[0004] Furthermore, while existing data augmentation methods such as SMOTE and GAN can expand the sample set, they fail to consider the physical constraints of the thermoplastic coating process, such as temperature safety thresholds and regulatory limits on plasticizer addition, potentially generating invalid data that does not meet production specifications. Deep learning models such as CNNs require extensive training data and are difficult to adapt to the small-batch, high-cost data characteristics of the pharmaceutical industry. Traditional regression models perform poorly when dealing with high-dimensional nonlinear relationships and class label prediction tasks, making them unable to meet the practical needs of thermoplastic coating process parameter optimization.

[0005] In view of this, the present invention is proposed. Summary of the Invention

[0006] In order to solve the above-mentioned technical problems existing in the prior art, the present invention provides a system and method for setting process parameters for the preparation of thermoplastic-coated controlled-release tablets, which solves the problems of quality fluctuations, overfitting of small samples, and difficulty in describing nonlinear relationships caused by the traditional heat-sealing process relying on manual experience, and realizes the accurate prediction and optimization of the production parameters of thermoplastic-coated controlled-release tablets, thereby improving production efficiency and quality stability.

[0007] To achieve the above object, the technical solution of the present invention is as follows: In a first aspect, a system for setting process parameters for preparing thermoplastic-coated controlled-release tablets comprises: Data processing module: used to collect parameter information data and construct it into a sample data set, and perform enhancement processing on the sample data set to obtain an enhanced data set; Model building module: used to build a prediction model with support vector machine as the core classifier, and define the core parameter space through kernel function; Model training module: used to standardize the enhanced data set of the data processing module and perform model training on the model building module to generate a prediction model with the highest accuracy parameter combination; Parameter setting module: used to use the prediction model of the parameter combination with the highest accuracy to predict the input processing parameter combination, and generate and output the optimal parameter combination based on the prediction result.

[0008] Furthermore, the data processing module includes: Raw data input unit: used to receive raw data sets including plasticizer content, heat sealing temperature, heat sealing time, pressure level and heat sealing effect level; Parameter perturbation unit: used to perturb continuous parameters based on process specifications and equipment safety thresholds, while discrete parameters remain unchanged; Invalid sample filtering unit: used to eliminate invalid samples that exceed the drug production specifications and expand the sample size of the data set to a preset multiple of the original sample.

[0009] Furthermore, the disturbance range of the continuous parameter is limited to: plasticizer content , heat sealing temperature , heat sealing time .

[0010] Furthermore, the model building module includes: Classifier selection unit: used to adopt support vector machine as the core classifier; Kernel function selection unit: used to select RBF kernel and linear kernel as backup kernel functions, and determine the optimal kernel function type and parameter range through grid search; Parameter space design unit: used to define the RBF kernel parameter space and linear kernel parameter space.

[0011] Furthermore, the RBF kernel parameter space is: ; The linear kernel parameter space is: .

[0012] Furthermore, the model training module includes: Data standardization unit: used to standardize the continuous process parameters of the enhanced data set of the data processing module; Stratified sampling unit: used to construct a stratified data set, and divide the enhanced data set normalized by the data normalization unit into a training set and a test set; Cross-validation unit: used to train the prediction model using a grid search strategy and three-fold cross-validation to obtain a prediction model with the highest accuracy parameter combination.

[0013] Furthermore, the continuous process parameters include: plasticizer content, heat sealing temperature and heat sealing time.

[0014] Furthermore, the formula for the normalization process is:

[0015] in, is the original eigenvalue, is the mean value of the feature in the training set, is the standard deviation of the feature in the training set.

[0016] Furthermore, the parameter setting module includes: Prediction result generation unit: used to receive the processing parameter combination input by the user and output the corresponding heat sealing effect level prediction result; Parameter screening unit: Generates multiple sets of candidate parameter combinations based on the prediction results and the plasticizer content minimization target; Human-computer interaction unit: displays parameter combinations and corresponding effects through a visual interface.

[0017] In a second aspect, a method for setting process parameters for preparing thermoplastic-coated controlled-release tablets comprises the following steps: S1. Data processing: Collect parameter information data and construct it into a sample data set, and perform enhancement processing on the sample data set to obtain an enhanced data set; S2. Model construction: Build a prediction model with support vector machine as the core classifier, and define the core parameter space through kernel function; S3, model training: normalizing the enhanced data set and training the prediction model to generate a prediction model with the highest accuracy parameter combination; S4. Parameter setting: using the prediction model of the parameter combination with the highest accuracy, predict the input processing parameter combination, and generate and output the optimal parameter combination based on the prediction result.

[0018] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a system and method for setting process parameters for the preparation of thermoplastic-coated controlled-release tablets. A data processing module collects parameter information and eliminates invalid samples to expand the data set. A model-building module uses a support vector machine as the core classifier to construct a parameter space containing both RBF and linear kernels. A model training module normalizes continuous parameters and combines grid search and three-fold cross-validation to generate a highly accurate prediction model. Based on the prediction results, a parameter-setting module generates candidate parameter combinations and visually outputs the optimal solution. This system achieves intelligent optimization of production parameters for thermoplastic-coated controlled-release tablets, significantly improving production stability and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 Schematic diagram of the architecture of the system for setting process parameters for preparing thermoplastic-coated controlled-release tablets provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The technical solution of the present invention will be clearly described below in conjunction with the accompanying drawings. Obviously, the described embodiments are not all embodiments of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0021] It should be noted that, unless otherwise specifically stated, the relative arrangements of components and steps, and numerical expressions set forth in these embodiments should not be construed as limiting the scope of the present invention.

[0022] The following description of exemplary embodiments is merely illustrative and is not intended to limit the present invention, its application, or use in any sense. Technologies, methods, and apparatus known to those skilled in the art may not be discussed in detail herein, but to the extent applicable, such technologies, methods, and apparatuses should be considered part of this specification.

[0023] Example 1 See Figure 1 , Figure 1This is an architecture diagram of a process parameter setting system for thermoplastic-coated controlled-release tablet preparation proposed in the present invention. The system comprises a data processing module, a model building module, a model training module, and a parameter setting module, which are sequentially connected to implement a parameter optimization process through data interaction. This system uses a process parameter setting method combining constrained data enhancement and an adaptive support vector machine model to address key issues such as coating film damage and abnormal drug release caused by unstable heat seal quality during the tablet coating process. Specifically, the system may include: A1. Data processing module: used to collect parameter information data and construct a sample data set, and perform enhancement processing on the sample data set to obtain an enhanced data set. Specifically, it may include: A11. Raw data input unit: used to receive raw data sets containing plasticizer content, heat sealing temperature, heat sealing time, pressure level, and heat sealing effect level; design a four-factor, three-level experiment (plasticizer content, heat sealing temperature, pressure, time) on thermoplastic coating equipment, and collect 60 sets of raw samples, including continuous parameters, discrete parameters, and heat sealing effect level labels.

[0024] A12, parameter perturbation unit: used to perturb continuous parameters based on process specifications and equipment safety thresholds, while discrete parameters remain unchanged; among them, continuous parameters include: plasticizer content , heat sealing temperature and heat sealing time Second.

[0025] A13. Invalid sample filtering unit: used to eliminate invalid samples that exceed the drug production specifications and expand the sample size of the data set to a preset multiple of the original sample.

[0026] A2. Model building module: This module is used to build a prediction model with the support vector machine as the core classifier, and defines the core parameter space through the kernel function. Specifically, it may include: A21, classifier selection unit: used to adopt support vector machine as the core classifier; A22, kernel function selection unit: used to select RBF kernel and linear kernel as backup kernel functions, and determine the optimal kernel function type and parameter range through grid search; A23, parameter space design unit: used to define the RBF kernel parameter space and the linear kernel parameter space. The RBF kernel parameter space is: ; The linear kernel parameter space is: .

[0027] A3. Model training module: used to standardize the enhanced data set of the data processing module and perform model training on the model building module to generate a prediction model with the highest accuracy parameter combination; specifically, it may include: A31, Data Standardization Unit: used to standardize the continuous process parameters of the enhanced data set of the data processing module; the continuous process parameters include: plasticizer content, heat sealing temperature and heat sealing time. The standardization formula is:

[0028] in, is the original eigenvalue, is the mean value of the feature in the training set, is the standard deviation of the feature in the training set.

[0029] A32, stratified sampling unit: used to construct a stratified data set, and divide the enhanced data set normalized by the data normalization unit into a training set and a test set; A33, cross validation unit: used to train the prediction model using a grid search strategy and three-fold cross validation to obtain a prediction model with the highest accuracy parameter combination.

[0030] A4. Parameter setting module: used to use the prediction model with the highest accuracy parameter combination to predict the input processing parameter combination, and generate and output the optimal parameter combination based on the prediction result; specifically, it may include: A41, prediction result generating unit: used to receive the processing parameter combination input by the user and output the corresponding heat sealing effect level prediction result; A42, parameter screening unit: based on the prediction results and the plasticizer content minimization target, generates multiple sets of candidate parameter combinations; A43, Human-computer interaction unit: Display parameter combinations and corresponding effects through a visual interface.

[0031] Example 2 The present invention proposes a method for setting process parameters for preparing thermoplastic-coated controlled-release tablets, which specifically comprises the following steps: S1. Data processing: Collect parameter information data and construct a sample data set, and perform enhancement processing on the sample data set to obtain an enhanced data set. The specific steps of data processing include: S11. Conduct experiments with different parameter combinations on the thermoplastic coating equipment installed in the laboratory. According to the experimental process, the integrity of the tablet coating is affected by four main variables: plasticizer content, heat sealing temperature, heat sealing pressure, and heat sealing time. During the experimental design phase, each variable was set to three levels according to the pre-set level range: plasticizer content was set to three levels, namely 15 microliters, 30 microliters, and 60 microliters; heat sealing temperature was set to 75 degrees Celsius, 85 degrees Celsius, and 95 degrees Celsius; heat sealing pressure was divided into three levels, namely low pressure, medium pressure, and high pressure; and heat sealing time was set to 1 second, 2 seconds, and 3 seconds.

[0032] As shown in Table 1, 81 processing parameter combinations were constructed using the full factorial approach. Sixty of these 81 parameter combinations were randomly selected for field experiments, and the heat sealing results generated by each parameter combination were used as the original sample.

[0033] Table 1 Factor level table

[0034] S12. To address the problem of a small number of original sample datasets and potential deviation in model training due to class imbalance in the heat seal grade classification task, the present invention performs data enhancement on the original sample dataset based on domain knowledge, expands the size of the original training dataset, and thereby avoids model overfitting and improves model robustness. Specifically, the present invention includes the following steps: S121. Input the original data set, which contains the heat sealing process parameters and quality inspection labels of thermoplastic-coated controlled-release tablets. These features include: Continuous parameter: Plasticizer content , heat sealing temperature , heat sealing time ; Discrete parameter: pressure level , where 1 is low pressure, 2 is medium pressure, and 3 is high pressure); Label: The effectiveness level of the heat sealing process (y∈{1,2,3,4}, 1 is excellent, 2 is suboptimal, 3 is qualified, and 4 is unqualified).

[0035] S122. Define the range of parameter perturbations based on the manufacturing process specifications of thermoplastic coated controlled-release tablets and the safety threshold of the equipment. Based on process experience and equipment capability, set the perturbation range of continuous parameters as follows: the perturbation range of plasticizer content is , the disturbance range of heat sealing temperature is , the disturbance range of heat sealing time is Since the pressure level is a discrete feature, no disturbance is performed. S123. Based on step 122, the original dataset is perturbed a preset number of times, which may be 5 times. After the perturbation is completed, invalid samples that exceed the pharmaceutical production specifications are eliminated, and the dataset sample size is ultimately expanded to a preset multiple of the original sample size, which may be 6 times. Thus, a new dataset is synthesized, effectively improving the robustness of model training.

[0036] S2. Model Construction: Build a prediction model using a support vector machine (SVM) as the core classifier, and define the core parameter space using a kernel function. This model first uses data augmentation before training, then combines it with a support vector machine (SVM) as the core classifier, improving model performance through parameter optimization. This allows for accurate predictions on small sample datasets. The specific steps for model construction include: S21. Core classifier selection. The support vector machine model is selected as the core classifier. The support vector machine has the advantage of processing small samples and high-dimensional data. It can improve the generalization ability of the model by maximizing the classification interval and is suitable for high-dimensional nonlinear classification tasks of heat sealing process parameters.

[0037] S22. Kernel function selection. Considering that there are both linear relationships such as the synergistic effect between temperature and time and nonlinear effects such as the interaction between plasticizer and temperature in the tablet coating process, the RBF kernel and the linear kernel are selected as backup kernel functions respectively. Subsequently, the grid search algorithm is used to select the kernel function that best suits the data characteristics.

[0038] S23, core parameter definition, determine the core parameter value range related to the kernel function, and set the value range of the kernel function kernel to , the value range of the regularization coefficient C is set to For the RBF kernel, the value range of its width gamma is set to Through different combinations of these parameters, the model performance can be optimized.

[0039] S24. Design parameter space. Design two independent subspaces, namely RBF kernel parameter space and linear kernel parameter space. The parameters in RBF kernel parameter space are: , corresponding to the search range of the RBF kernel function; The parameters contained in the linear kernel parameter space are: , which corresponds to the search range of the linear kernel function. Retaining two kernel types enables the system to automatically select the decision boundary form that best suits the data characteristics, taking into account the classification requirements of different types of data relationships.

[0040] S3. Model training: Standardize the enhanced dataset of the data processing module and train the prediction model using a grid search strategy and three-fold cross-validation to generate a prediction model with the highest accuracy parameter combination. After the prediction model is built, it needs to be trained. The specific steps of model training are as follows: S31. Standardize the continuous process parameters in the data set generated after data enhancement, wherein the process parameters are plasticizer content, temperature, and time. The standardization formula is as follows:

[0041] in, is the original eigenvalue, is the mean value of the feature in the training set, is the standard deviation of the feature in the training set. This step can eliminate dimensional differences and make all features follow a standard normal distribution with a mean of 0 and a variance of 1, thus improving the robustness of the SVM to feature scale sensitivity.

[0042] S32. Construct a stratified dataset and use stratified sampling to divide the data into a training set and a test set, with 80% of the data used as the training set and 20% as the test set. During the partitioning process, ensure that the distribution of labels in the training and test sets is consistent to avoid data bias that may affect model evaluation.

[0043] S33. A grid search strategy is used to traverse 12 parameter combinations, 9 of which are RBF kernel parameter combinations and 3 are linear kernel parameter combinations. Combined with three-fold cross-validation, the training set is divided into 3 subsets to ensure that the proportion of samples of each category in each data subset is consistent with the original data. For each parameter combination, 2 subsets are used for training in turn, and the remaining 1 subset is used for validation, and the average performance is finally calculated. It should be noted that three-fold cross-validation is only performed within the training set, and the test set does not participate in parameter selection throughout the process to ensure the objectivity of the model evaluation. In this way, the parameter combination with the highest cross-validation accuracy is selected.

[0044] S34, optimal model generation, after obtaining the best parameter combination, automatically input it into the model. of Function, automatically uses the best parameters to train on the complete training set, without manual retraining, and finally determines the model used for prediction, ensuring that the model has high generalization ability and prediction accuracy S4. Parameter setting: Use the prediction model with the highest accuracy parameter combination to predict the input processing parameter combination, and generate and output the optimal parameter combination based on the prediction results. Specifically, this may include: S41. Parameter summary: In the user interface, the user enters all parameter combinations, and the system calculates the processing results corresponding to each parameter combination. All processing parameter combinations and corresponding processing results are combined into a summary table, which allows the user to intuitively view the processing effects under different parameter combinations and provides a clear data reference for parameter selection. As shown in Table 2: Table 2 Summary table example

[0045] S42. Parameter Selection: When selecting processing parameters, follow the principle of multi-objective optimization. While ensuring excellent processing results, specifically, prioritizing heat seal levels 1 and 2, choose processing parameter combinations with low plasticizer content. This not only reduces production costs but also improves processing efficiency, achieving a dual optimization of quality and cost.

[0046] S43. Parameter Setting: Before operating the thermoplastic coating machine, first enter the selected processing parameters into the machine control panel to ensure that the parameters are entered correctly. Then, start the machine for production according to the equipment operating procedures. Through standardized parameter settings, the consistency of production between batches is guaranteed, thereby improving the efficiency of industrial production and the stability of product quality. In summary, the present invention has the following advantages: 1. Eliminate subjective judgment by operators, avoid batch-to-batch parameter deviations caused by manual adjustments, ensure the stability and consistency of the production quality of thermoplastic-coated controlled-release tablets from the source, and solve the problem of quality fluctuations caused by reliance on manual experience; 2. Accurately predict the heat sealing effect level of different parameter combinations before production, solving the problem of resource waste caused by the inability to predict the effect of process parameters in advance; 3. Significantly improve model robustness in small sample scenarios, avoid overfitting, and solve the problem that existing linear modeling cannot describe complex coupling relationships.

[0047] The above specific embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to examples, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the scope of the technical solutions of the present invention, and all of these should be included in the scope of the claims of the present invention.

Claims

1. A system for setting process parameters for the preparation of thermoplastic coated controlled-release tablets, characterized in that: include: Data processing module: used to collect parameter information data and construct it into a sample data set, and perform enhancement processing on the sample data set to obtain an enhanced data set; Model building module: used to build a prediction model with support vector machine as the core classifier, and define the core parameter space through kernel function; Model training module: used to standardize the enhanced data set of the data processing module and perform model training on the model building module to generate a prediction model with the highest accuracy parameter combination; Parameter setting module: used to use the prediction model of the parameter combination with the highest accuracy to predict the input processing parameter combination, and generate and output the optimal parameter combination based on the prediction result.

2. The system for setting process parameters for preparing thermoplastic coated controlled-release tablets according to claim 1, characterized in that: The data processing module includes: Raw data input unit: used to receive raw data sets including plasticizer content, heat sealing temperature, heat sealing time, pressure level and heat sealing effect level; Parameter perturbation unit: used to perturb continuous parameters based on process specifications and equipment safety thresholds, while discrete parameters remain unchanged; Invalid sample filtering unit: used to eliminate invalid samples that exceed the drug production specifications and expand the sample size of the data set to a preset multiple of the original sample.

3. The system for setting process parameters for preparing thermoplastic coated controlled-release tablets according to claim 2, characterized in that: The disturbance range of the continuous parameter is limited to: plasticizer content , heat sealing temperature , heat sealing time .

4. The system for setting process parameters for preparing thermoplastic coated controlled-release tablets according to claim 1, characterized in that: The model building module includes: Classifier selection unit: used to adopt support vector machine as the core classifier; Kernel function selection unit: used to select RBF kernel and linear kernel as backup kernel functions, and determine the optimal kernel function type and parameter range through grid search; Parameter space design unit: used to define the RBF kernel parameter space and linear kernel parameter space.

5. The system for setting process parameters for preparing thermoplastic coated controlled-release tablets according to claim 4, characterized in that: The RBF kernel parameter space is: ; The linear kernel parameter space is: .

6. The system for setting process parameters for preparing thermoplastic coated controlled-release tablets according to claim 1, characterized in that: The model training module includes: Data standardization unit: used to standardize the continuous process parameters of the enhanced data set of the data processing module; Stratified sampling unit: used to construct a stratified data set, and divide the enhanced data set normalized by the data normalization unit into a training set and a test set; Cross-validation unit: used to train the prediction model using a grid search strategy and three-fold cross-validation to obtain a prediction model with the highest accuracy parameter combination.

7. The system for setting process parameters for preparing thermoplastic coated controlled-release tablets according to claim 6, characterized in that: The continuous process parameters include: plasticizer content, heat sealing temperature and heat sealing time.

8. The system for setting process parameters for preparing thermoplastic coated controlled-release tablets according to claim 6, characterized in that: The formula for the standardization process is: in, is the original eigenvalue, is the mean value of the feature in the training set, is the standard deviation of the feature in the training set.

9. The system for setting process parameters for preparing thermoplastic coated controlled-release tablets according to claim 1, characterized in that: The parameter setting module includes: Prediction result generation unit: used to receive the processing parameter combination input by the user and output the corresponding heat sealing effect level prediction result; Parameter screening unit: Generates multiple sets of candidate parameter combinations based on the prediction results and the plasticizer content minimization target; Human-computer interaction unit: displays parameter combinations and corresponding effects through a visual interface.

10. A method for setting process parameters for preparing thermoplastic coated controlled-release tablets, characterized in that: The specific steps include: S1. Data processing: Collect parameter information data and construct it into a sample data set, and perform enhancement processing on the sample data set to obtain an enhanced data set; S2. Model construction: Build a prediction model with support vector machine as the core classifier, and define the core parameter space through kernel function; S3, model training: normalizing the enhanced data set and training the prediction model to generate a prediction model with the highest accuracy parameter combination; S4. Parameter setting: using the prediction model of the parameter combination with the highest accuracy, predict the input processing parameter combination, and generate and output the optimal parameter combination based on the prediction result.

Images