Mouth-dissolving film preparation equipment for medicines and health care products

Abstract

The invention belongs to the technical field of medical health care product production equipment, and discloses film preparation equipment of an oral soluble film for medicines and health care products. The equipment sequentially comprises a PET (polyethylene terephthalate) unwinding module, a roller passing module, a tension module, a coating module, a thickness detection module, a thermostat with a built-in infrared dryer, a deviation correction module, a CCD (charge coupled device) detection module, a separation module, a tension swing rod and a double-winding module along a film transmission path. And the coating module, the thickness detection module and the drying module form a closed-loop control system. The method comprises the steps of tension unwinding, coating thickness closed-loop adjustment, infrared precuring and gradient drying, multispectral online detection and membrane separation winding. The innovation points are as follows: (1) multi-spectrum (visible light / NIR / Raman) and AI defect classification are integrated, and 50 [mu] m grade traditional Chinese medicine component distribution detection is realized; (2) the error of the LSTM dissolution rate prediction model is less than or equal to + / -4.5%; and (3) the edge calculation unit outputs a quality score Q in real time through a random forest algorithm to trigger process dynamic adjustment. The bottleneck of mass production of the traditional Chinese medicine oral soluble film is broken through, and the yield is increased by 40%.

Description

Technical Field

[0001] The invention belongs to the technical field of pharmaceutical and health-care product production equipment, and specifically discloses a film-making device for orally dissolving films for pharmaceuticals and health-care products. Background Art

[0002] As a new drug delivery system, oral dissolving films for pharmaceuticals offer significant advantages in improving patient compliance and are particularly suitable for children, the elderly, and those with swallowing disorders. In recent years, oral dissolving films for traditional Chinese medicine have attracted considerable attention due to their minimal side effects and stable efficacy. However, their industrialization faces three technical bottlenecks:

[0003] 1. Equipment dependence on imports: High-end film-making equipment has long been monopolized by foreign countries, and domestic equipment has gaps in key indicators such as coating accuracy (especially high-viscosity Chinese herbal extracts) and drying uniformity;

[0004] 2. Challenges in Traditional Chinese Medicine (TCM) quality control: The coexistence of multiple components in compound TCMs makes it impossible to monitor drug distribution uniformity online through traditional visual inspection, and there is a lack of effective control measures for the drying and degradation of active ingredients.

[0005] 3. Process-efficacy disconnect: Existing equipment is unable to establish a quantitative correlation between production process parameters (such as film thickness and drying temperature) and efficacy characteristics (such as dissolution rate), and relies on offline sampling, which restricts quality stability.

[0006] This industry has the following demand pain points:

[0007] There is an urgent need to develop domestic equipment for orally dissolving films with independent intellectual property rights, and to break through core technologies such as coating uniformity control and online monitoring of Chinese medicine components; there is an urgent need to build an intelligent quality control system to achieve industrial upgrading from "trial and error" to "data-driven"; it is necessary to comply with the trend of green manufacturing and develop environmentally friendly processes to reduce production energy consumption and waste emissions. Summary of the Invention

[0008] In response to the above problems, the present invention discloses an orally disintegrating film-making device for medicines and health products and a method for using the same. The device integrates precision coating control, multi-spectral intelligent detection, and a drug efficacy prediction model into a high-end orally disintegrating film-making device and method, aiming to achieve high-quality, efficient, and green manufacturing of orally disintegrating films for Chinese medicine compounds.

[0009] The purpose of the present invention is achieved through the following technical solutions.

[0010] An orally dissolving film making device for medicine and health care products comprises the following components in order along the film transmission path:

[0011] PET film unwinding module, used to release the PET substrate film;

[0012] The PET film roller module connected to the output end of the unwinding module is used to guide the film to turn;

[0013] The tension module connected to the roller module detects and adjusts the film tension in real time;

[0014] The coating module is located downstream of the tension module and applies the pharmaceutical slurry to the surface of the PET film;

[0015] The thickness detection module is located next to the coating module outlet and monitors the wet film thickness online;

[0016] Constant temperature box module, receiving the coated film and performing main drying;

[0017] The infrared lamp drying module built into the constant temperature box module pre-cures the wet film;

[0018] The process correction module is located at the exit of the constant temperature box module to correct the lateral deviation of the film;

[0019] The surface defect CCD detection module is set downstream of the correction module to image the dry film surface;

[0020] Medical film separation module, which physically separates the PET substrate from the cured medical film;

[0021] The tension swing arm module is connected to the medical film output end of the separation module to adjust the winding tension;

[0022] The medical film winding module receives the finished medical film after tension adjustment;

[0023] PET film winding module, independently receiving the separated PET substrate;

[0024] Among them, the coating module, thickness detection module, and constant temperature box module constitute a closed-loop control system, which adjusts the coating parameters in real time according to thickness feedback.

[0025] Furthermore, the above-mentioned orally dissolving film making equipment for medicine and health care products,

[0026] The coating module adopts a slit extrusion coating head and is equipped with a high-precision servo motor to drive the scraper, and the scraper gap adjustment accuracy is ≤1μm; a constant temperature flow channel is provided inside the coating head, and the temperature control range is 30–80℃±0.5℃.

[0027] Furthermore, in the above-mentioned orally dissolving film making equipment for medicine and health products, the tension module and the tension swing arm module are integrated with a multi-stage magnetic powder brake, combined with a laser displacement sensor to detect the film deformation in real time, and a dynamic tension compensation signal is output through a PID algorithm to control the film tension fluctuation within ±0.5N.

[0028] The present invention discloses a film-making method based on the above-mentioned device, comprising the following steps:

[0029] S1.PET film is unwound and maintained at a constant tension by the tension module;

[0030] S2. The coating module evenly coats the pharmaceutical slurry onto the PET film to a thickness of 10–200 μm.

[0031] S3. The thickness detection module feeds back data to the coating module in real time, and the coating parameters are adjusted in a closed loop;

[0032] S4. The wet film is pre-cured in an infrared lamp drying module and then dried in a constant temperature chamber module. The drying temperature is adjustable from 40°C to 120°C.

[0033] S5. Surface defect CCD detection module performs online visual inspection of dry film;

[0034] S6. The medical film separation module peels off the PET carrier and the medical film is rolled up independently.

[0035] Furthermore, the above film-making method is characterized in that:

[0036] In step S4, the infrared drying band is 2.5–5 μm, and the power density can be adjusted in the range of 1–10 kW / m2; the constant temperature box adopts zoned and segmented temperature control, and each temperature zone is independently PID-regulated, with a temperature difference of ≤±1°C.

[0037] Furthermore, in the above film forming method, in step S5, the surface defect CCD detection module is integrated with a multispectral imaging system, including:

[0038] a) Visible light CCD camera to detect bubbles, scratches and dimensional deviations;

[0039] b) Near infrared spectrometer, real-time scanning of the distribution uniformity of active pharmaceutical ingredients;

[0040] c) Raman spectroscopy probe, online identification of the content of characteristic components of traditional Chinese medicine;

[0041] Based on the fusion of multi-source data, a drug distribution heat map is generated with a spatial resolution of ≤50μm.

[0042] Furthermore, in the above film forming method, the surface defect CCD detection module is further provided with d) an AI defect classification subsystem, including:

[0043] Pre-trained 3D convolutional neural network model, including input layer feature extraction layer;

[0044] The input layer receives a multispectral image sequence; the feature extraction layer uses the ResNet-50 architecture to output defect types and confidence levels; the output defect types include bubbles, cracks, and impurities;

[0045] When the confidence level is >95%, a real-time alarm is triggered and the defect location coordinates are synchronously marked in the production log.

[0046] Furthermore, in the above film-forming method, in step S5, e) a dissolution rate prediction model is established:

[0047] The following operations are performed by the e) dissolution rate prediction model in the surface defect CCD detection module:

[0048] Real-time collection of film thickness data, drug distribution thermogram and drying temperature curve from the thickness detection module;

[0049] The data is input into the LSTM neural network, which outputs the predicted dissolution time T with an error of ≤±5%;

[0050] If T exceeds the set threshold time, the coating thickness or drying parameters of subsequent batches will be automatically adjusted.

[0051] Furthermore, in the above film-making method, the dissolution rate prediction model is optimized by transfer learning:

[0052] The initial model was trained based on laboratory dissolution test data, with a dataset of ≥1000 groups;

[0053] After online deployment, the actual dissolution data of each batch is fed back to the cloud server;

[0054] Adopting a federated learning framework, model weights are updated across devices to improve prediction generalization.

[0055] Furthermore, the above film making method further includes an edge computing quality decision unit for:

[0056] Integrate real-time detection data of thickness detection module and surface defect CCD detection module;

[0057] Run the lightweight random forest algorithm to calculate the quality score Q, where the value range of Q is 0–100;

[0058] If Q<90, the coating module will be linked to adjust the feeding rate or the drying module will adjust the temperature;

[0059] If Q<70, the winding module is triggered to stop and the defective area is marked.

[0060] Compared with the existing technology, the present invention has the following advantages and beneficial effects:

[0061] 1. Use synchronous unwinding and rewinding, and tension control technology to make the PET film run at a constant speed, ensuring that the PET film runs at a constant speed, thereby ensuring the uniformity of the thickness of the pharmaceutical film;

[0062] 2. Adopt thickness detection technology and servo-controlled medical film thickness technology to achieve closed-loop control. When it is detected that the thickness does not meet the requirements, the coating module can be automatically adjusted by servo to achieve the qualified thickness.

[0063] 3. Using infrared lamp drying technology and constant temperature box heating and baking technology to bake and dry the wet film, which can quickly solidify the wet film and improve production efficiency;

[0064] 4. Use CCD defect detection after film formation to detect surface defects and dimensions of the film. If any defect is found, an alarm will be issued in time to reduce the waste of raw materials;

[0065] 5. Green manufacturing process technology route: Select biodegradable and environmentally friendly materials, optimize production processes, reduce energy consumption and waste emissions, and promote the sustainable development of orally dissolving pharmaceutical film manufacturing through process optimization and equipment improvement.

[0066] 6. Furthermore, in some embodiments, the present invention also innovatively integrates multi-spectral intelligent detection and efficacy prediction models to achieve high-sensitivity online monitoring of the distribution of Chinese medicine components and membrane defects, and optimizes process parameters in real time through an edge computing system, significantly improving product consistency and production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 A schematic diagram of the film-making equipment of the present invention;

[0068] Among them: 1. PET film unwinding module; 2. PET film roller module; 3. Tension module; 4. Coating module; 5. Thickness detection module; 6. Constant temperature box module; 7. Infrared lamp drying module; 8. Process correction module; 9. Film surface defect CCD detection module; 10. Pharmaceutical film separation module; 11. Tension swing arm module; 12. Pharmaceutical film winding module; 13. PET film winding module. DETAILED DESCRIPTION

[0069] In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention is further described in detail below. However, it should be understood that the specific embodiments described herein are merely for the purpose of explaining the present invention and are not intended to limit the scope of the invention. In addition, in the following description, the description of known structures and technologies is omitted to avoid unnecessary confusion of the concept of the present invention. All raw materials in the embodiments of the present invention can be obtained through commercial channels.

[0070] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present invention will be described in detail below with reference to the embodiments.

[0071] Those skilled in the art will understand that:

[0072] The detection of characteristic components of traditional Chinese medicine requires the establishment of a standard spectral library of target components in advance;

[0073] The dataset used for machine learning model training should comply with the GB / T 19001 quality management standard;

[0074] The adjustment range of equipment parameters is limited by the mechanical structure, and the specific thresholds need to be determined through process verification.

[0075] Example 1

[0076] like Figure 1 The device for forming orally dissolving films for pharmaceutical and health products includes the following steps along the film transmission path:

[0077] PET film unwinding module 1, used for releasing PET substrate film;

[0078] The PET film roller module 2 connected to the output end of the unwinding module 1 is used to guide the film to turn;

[0079] The tension module 3 connected to the roller module 2 detects and adjusts the film tension in real time;

[0080] The coating module 4 is arranged downstream of the tension module 3 and applies the pharmaceutical slurry to the surface of the PET film;

[0081] The thickness detection module 5 is located next to the outlet of the coating module 4 and monitors the wet film thickness online;

[0082] Constant temperature box module 6, receives the coated film and performs main drying;

[0083] The infrared lamp drying module 7 built into the constant temperature box module 6 pre-cures the wet film;

[0084] The process deviation correction module 8 is located at the exit of the constant temperature box module 6 and corrects the lateral deviation of the film;

[0085] The surface defect CCD detection module 9 is arranged downstream of the correction module 8 to image the dry film surface;

[0086] The medical film separation module 10 physically separates the PET substrate from the solidified medical film;

[0087] The tension swing arm module 11 is connected to the medical film output end of the separation module 10 to adjust the winding tension;

[0088] The medical film winding module 12 receives the finished medical film after tension adjustment;

[0089] The PET film winding module 13 independently receives the separated PET substrate;

[0090] Among them, the coating module 4, the thickness detection module 5, and the constant temperature box module 6 constitute a closed-loop control system, and the coating parameters are adjusted in real time according to the thickness feedback.

[0091] Preferably, the coating module 4 adopts a slit extrusion coating head equipped with a high-precision servo motor-driven scraper with a scraper gap adjustment accuracy of ≤1μm; a constant temperature flow channel is provided inside the coating head with a temperature control range of 30–80℃±0.5℃.

[0092] Preferably, the tension module 3 and the tension rocker module 11 are integrated with a multi-stage magnetic powder brake, combined with a laser displacement sensor to detect the film deformation in real time, and output a dynamic tension compensation signal through a PID algorithm to control the film tension fluctuation within ±0.5N.

[0093] The film-making method of the above-mentioned orally disintegrating film-making equipment for medicines and health products generally comprises the following steps:

[0094] S1. PET film is unwound and maintained at a constant tension by tension module 3;

[0095] S2. Coating module 4 evenly coats the pharmaceutical slurry on the PET film to a coating thickness of 10–200 μm;

[0096] S3. The thickness detection module 5 provides real-time feedback data to the coating module 4, and closes the loop to adjust the coating parameters;

[0097] S4. The wet film is pre-cured by the infrared lamp drying module 7 and then dried in the constant temperature box module 6. The drying temperature is adjustable in a gradient of 40–120°C.

[0098] S5. Surface defect CCD detection module 9 performs online visual inspection of the dry film;

[0099] S6. The medical film separation module 10 peels off the PET carrier, and the medical film is independently rolled up.

[0100] Preferably, the above film-making method specifically comprises the following steps:

[0101] First, thread the PET film onto the tape according to the tape routing diagram.

[0102] During operation, the PET film unwinding module 1 unwinds at a constant speed, the PET film roller module 2 provides support for the PET film, and the tension module 3 detects the tension of the PET film in real time;

[0103] The coating module 4 can realize the uniform coating of the medicine and health care product slurry onto the PET film. The coating thickness, width and uniformity can be set and adjusted in the coating module. A very thin layer of slurry is transported forward along with the PET film.

[0104] The thickness detection module 5 performs real-time thickness detection on the coated pharmaceutical or health product film. If the thickness exceeds the specification, it can automatically adjust or alarm.

[0105] The coated medical film adheres to the PET film and enters the constant temperature box module 6 along with the PET film. The constant temperature box realizes the function of heating and drying the medical film, so that the medical film in the glue state is quickly dried into a solid film.

[0106] The dried medical film is attached to the PET film and leaves the constant temperature box module 6. It is corrected by the process correction module 8 to ensure that the previous deviation can be corrected.

[0107] The PET film and medical film after the route correction enter the surface defect CCD detection module 9, which takes a photo of the surface of the medical film to detect whether there are defects such as bubbles and poor coating size on the surface;

[0108] After the CCD inspection, the film enters the pharmaceutical film separation module 10, which is used to separate the PET film and the pharmaceutical film to prepare for separate rolls.

[0109] After separation, the PET film can be directly connected to the PET film winding module 13 for winding, and the pharmaceutical film passes through the tension swing rod mechanism 11 to adjust the tension of the pharmaceutical film again, and is wound with very small tension, and finally enters the pharmaceutical film winding module 12.

[0110] Example 2

[0111] Multispectral detection and AI defect classification system

[0112] Perform equipment upgrade based on Example 1:

[0113] Integrate a multispectral imaging system into the surface defect CCD detection module (9):

[0114] a) Visible light CCD camera (wavelength 400-700 nm): Scans the film surface at 100 frames per second with a resolution of 10 μm / pixel, detecting bubbles (diameter > 50 μm), scratches (length > 100 μm), and dimensional deviations (width error ± 5%).

[0115] b) Near-infrared spectrometer (900-1700 nm): Perform a line scan every 5 cm of film length and analyze the distribution uniformity of active pharmaceutical ingredients (e.g., tanshinone) using OPUS software. The standard deviation must be <3%;

[0116] c) Raman spectroscopy probe (785nm excitation light source): Focus diameter 20μm, sample 10×10 grid points on the membrane surface, match the characteristic peaks of traditional Chinese medicine (such as berberine at 1600cm -1When the content deviation exceeds ±5%, an early warning is triggered.

[0117] AI defect classification operation process:

[0118] 1. Data fusion: Input multispectral data into a pre-trained 3D-CNN model (TensorFlow framework, input layer size 256×256×4, including visible light + 3 NIR feature bands);

[0119] 2. Defect identification:

[0120] The feature extraction layer uses ResNet-50 and outputs a 1024-dimensional feature vector;

[0121] 3. The fully connected layer classifies the defect type (bubble / crack / impurity), and Softmax outputs the confidence level;

[0122] Execution mechanism: When the confidence level is >95%, the PLC controls the alarm to sound, and at the same time marks the defect coordinates in the production log (format: X:120mm, Y:35mm) and saves the defect image.

[0123] 4. Training data: The model was trained based on 50,000 labeled membrane defect samples (including simulated bubbles, cracks, and Chinese medicine impurities), with a test set accuracy of 98.2%.

[0124] Example 3

[0125] Dissolution rate prediction and federated learning optimization

[0126] Perform equipment upgrade based on Example 2:

[0127] Model construction (deployed on the industrial computer of the CCD detection module (9)):

[0128] 1. Input data:

[0129] Real-time thickness value of thickness detection module (5) (10 sampling points per second);

[0130] Drug distribution heat map (from Example 2);

[0131] Drying temperature curve (thermocouple data of 6 temperature zones in the constant temperature box);

[0132] 2. LSTM network structure:

[0133] Input layer 3 channels (thickness / distribution index / temperature), time step 60s;

[0134] Hidden layer with 128 neurons and Dropout = 0.2;

[0135] The output layer predicts the dissolution time T (unit: seconds);

[0136] 3. Closed loop control: If T>30s (set threshold), the feeding rate of the coating module (4) is automatically reduced by 5% or the drying temperature is increased by 10°C.

[0137] Federated learning optimization process:

[0138] Initial training: Based on 1,200 sets of laboratory dissolution test data (film thickness 100-200 μm, temperature 40-120°C, actual T value detected by dissolution instrument) (using tanshinone orally dissolving film as sample);

[0139] Online Update:

[0140] After each production batch is completed, upload the actual dissolution data (dissolution time of the membrane sample in 37°C pure water)

[0141] to the cloud;

[0142] The FedAvg algorithm is used to aggregate the weights of the three devices, and the encrypted communication protocol is AES-256;

[0143] The global model is updated monthly, and the forecast error has been reduced from ±8% to ±4.5%.

[0144] Example 4

[0145] Edge computing quality decision system

[0146] Perform equipment upgrade based on Example 2:

[0147] Hardware configuration:

[0148] The edge computing unit (NVIDIA Jetson Xavier NX) is installed in the electric control cabinet and connected to the thickness detection module (5) and the CCD detection module (9) via the EtherCAT bus.

[0149] Decision-making process:

[0150] Data input (refresh every 200ms):

[0151] Thickness uniformity index (CV value ≤ 3%);

[0152] Defect detection results (from Example 2);

[0153] standard deviation of drug distribution (from NIR spectroscopy);

[0154] Random Forest Algorithm:

[0155] Input features: thickness CV value, number of defects, standard deviation of drug distribution;

[0156] Output: quality score Q (0-100), Q=90 corresponds to the national standard passing line;

[0157] Real-time control:

[0158] If 85≤Q<90: reduce the coating speed by 10% and increase the temperature of the thermostat by 5°C;

[0159] If Q < 70: the winding module is triggered to stop suddenly, an audible and visual alarm is sounded, and the defective interval is recorded (such as "roll number 2025-Batch3, position 15.2m-15.5m") for the slitting machine to remove.

[0160] Model validation: Q value evaluation was performed on 200 batches of samples (samples containing multiple active ingredients of traditional Chinese medicine), and the correlation R with the manual quality inspection results was 2 =0.93.

[0161] As can be seen from the above examples, the present invention provides a high-precision, intelligent orally disintegrating film-making device and method for pharmaceutical and health products, which achieves high-quality mass production of orally disintegrating films for traditional Chinese medicine compounds through a four-level innovative architecture:

[0162] Basic manufacturing layer (Example 1)

[0163] The first integrated equipment architecture of "coating-thickness closed loop-segmented drying" solves the problem of film thickness uniformity (fluctuation ≤ ±1.5μm) through the linkage control of the tension module 3, thickness detection module 5 and constant temperature box (6);

[0164] The gradient drying process of infrared pre-curing 7 + hot air main drying 6 is adopted to ensure that the retention rate of active ingredients of traditional Chinese medicine is greater than 95%.

[0165] Intelligent detection layer (Example 2)

[0166] For the first time, "visible light + NIR + Raman" multispectral online detection is integrated into an orally dissolving film device, with a spatial resolution of 50μm;

[0167] The 3D-CNN-based AI defect classification system enables real-time identification of bubbles / cracks / traditional Chinese medicine impurities (with an accuracy rate of 98.2%), increasing defect detection efficiency by 10 times.

[0168] Drug efficacy prediction layer (Example 3)

[0169] Universities use the dissolution rate LSTM prediction model to correlate process parameters (thickness, temperature, component distribution) with drug efficacy, with a prediction error of ≤±4.5%;

[0170] Innovatively apply the federated learning framework, iteratively optimize models across devices, and promote the pharmaceutical process from "experience-driven" to "data-driven".

[0171] Quality Decision Layer (Example 4)

[0172] The edge computing unit runs a lightweight random forest algorithm and dynamically outputs a quality score Q (R2 = 0.93);

[0173] Establish a Q value linkage control mechanism to achieve millisecond response from detection to process adjustment (delay

[0174] <200ms), the scrap rate is reduced by 40%.

[0175] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, based on the innovative concept of the present invention, changes and modifications to the embodiments described herein, or equivalent structural or equivalent process transformations made using the contents of the present invention specification, directly or indirectly applying the above technical solutions to other related technical fields are all included in the scope of protection of the patent of the present invention.

Claims

1. A device for making orally dissolving films for medicines and health products, characterized in that: The film transmission path includes: A PET film unwinding module (1) for releasing the PET substrate film; A PET film roller module (2) connected to the output end of the unwinding module (1) is used to guide the film to turn; A tension module (3) connected to the roller module (2) detects and adjusts the film tension in real time; A coating module (4) is provided downstream of the tension module (3) and is used to coat the pharmaceutical slurry on the surface of the PET film; A thickness detection module (5), located adjacent to the outlet of the coating module (4), monitors the wet film thickness online; A constant temperature box module (6) receives the coated film and performs main drying; An infrared lamp drying module (7) built into the constant temperature box module (6) pre-cures the wet film; A process deviation correction module (8), located at the outlet of the constant temperature box module (6), corrects the lateral deviation of the film; A surface defect CCD detection module (9) is provided downstream of the deviation correction module (8) to image the dry film surface; A pharmaceutical film separation module (10) is used to physically separate the PET substrate from the solidified pharmaceutical film; A tension swing rod module (11) is connected to the medical film output end of the separation module (10) to adjust the winding tension; A medical film winding module (12) receives the tension-adjusted finished medical film; A PET film winding module (13) independently receives the separated PET substrate; The coating module (4), the thickness detection module (5), and the constant temperature box module (6) constitute a closed-loop control system, and the coating parameters are adjusted in real time according to the thickness feedback.

2. The orally dissolving film-making device for medicine and health care products according to claim 1, characterized in that: The coating module (4) adopts a slit extrusion coating head and is equipped with a high-precision servo motor to drive a scraper, and the scraper gap adjustment accuracy is ≤1μm; a constant temperature flow channel is provided inside the coating head, and the temperature control range is 30-80℃±0.5℃.

3. The film-making equipment according to claim 1, characterized in that: The tension module (3) and the tension swing arm module (11) are integrated with a multi-stage magnetic powder brake, and are combined with a laser displacement sensor to detect the film deformation in real time, and a dynamic tension compensation signal is output through a PID algorithm to control the film tension fluctuation within ±0.5N.

4. A film-making method based on the device according to any one of claims 1 to 3, characterized in that: The following steps are involved: S1. The PET film is unwound and maintained at a constant tension by the tension module (3); S2. Coating module (4) evenly coats the pharmaceutical slurry on the PET film with a coating thickness of 10–200 μm; S3. The thickness detection module (5) feeds back data to the coating module (4) in real time, and adjusts the coating parameters in a closed loop; S4. The wet film is pre-cured in the infrared lamp drying module (7) and then dried in the constant temperature box module (6). The drying temperature is adjustable in a gradient of 40–120°C. S5. Surface defect CCD detection module (9) performs online visual inspection of the dry film; S6. The medical film separation module (10) peels off the PET carrier and the medical film is independently rolled up.

5. The film forming method according to claim 4, wherein: In step S4, the infrared drying wavelength is 2.5–5 μm, and the power density can be adjusted in the range of 1–10 kW / m 2 The constant temperature box adopts zoned and segmented temperature control, and each temperature zone has independent PID adjustment, and the temperature difference is ≤±1℃.

6. The method according to claim 4, characterized in that: In step S5, the surface defect CCD detection module (9) is integrated with a multispectral imaging system, including: a) Visible light CCD camera (wavelength 400–700 nm) to detect bubbles, scratches, and dimensional deviations; b) Near-infrared spectrometer (900–1700 nm), real-time scanning of the distribution uniformity of active pharmaceutical ingredients; c) Raman spectroscopy probe (785nm / 1064nm), online identification of the content of characteristic components of traditional Chinese medicine; Based on the fusion of multi-source data, a drug distribution heat map is generated with a spatial resolution of ≤50μm.

7. The method according to claim 6, characterized in that: The surface defect CCD detection module (9) is further provided with d) an AI defect classification subsystem, comprising: A pre-trained 3D convolutional neural network (CNN) model, including an input layer and a feature extraction layer; The input layer receives a multispectral image sequence; the feature extraction layer uses the ResNet-50 architecture to output defect types and confidence levels; the output defect types include bubbles, cracks, and impurities; When the confidence level is >95%, a real-time alarm is triggered and the defect location coordinates are synchronously marked in the production log.

8. The method according to claim 4, wherein: In step S5, e) a dissolution rate prediction model is also established: The following operations are performed by the e) dissolution rate prediction model in the surface defect CCD detection module (9): Real-time collection of film thickness data, drug distribution thermodynamic map and drying temperature curve of the thickness detection module (5); The data is input into the LSTM neural network, which outputs the predicted dissolution time T with an error of ≤±5%; If T exceeds the set threshold time, the coating thickness or drying parameters of subsequent batches will be automatically adjusted.

9. The method according to claim 8, characterized in that: The dissolution rate prediction model was optimized through transfer learning: The initial model was trained based on laboratory dissolution test data, with a dataset of ≥1000 groups; After online deployment, the actual dissolution data of each batch is fed back to the cloud server; Adopting a federated learning framework, model weights are updated across devices to improve prediction generalization.

10. The method according to claim 4, characterized in that: It also includes an edge computing quality decision unit for: Integrating real-time detection data of the thickness detection module (5) and the surface defect CCD detection module (9); Run the lightweight random forest algorithm to calculate the quality score Q, where the value range of Q is 0–100; If Q<90, the coating module (4) is linked to adjust the feeding rate or the drying module (6) is linked to adjust the temperature; If Q<70, the winding module is triggered to stop and the defective area is marked.

Images