Preparation method for evodia rutaecarpa formula granule formulation and system thereof

Abstract

Disclosed in the present invention are a preparation method for an evodia rutaecarpa formula granular formulation and a system thereof. The preparation method comprises: adding water into evodia rutaecarpa decoction pieces for the first time to decoct to obtain a first decoction medicine solution and medicine residues; adding water into the medicine residues for the second time to decoct to obtain a second decoction medicine solution; mixing the first decoction medicine solution and the second decoction medicine solution, and performing vacuum concentration to obtain a concentrated solution, wherein the relative density of the concentrated solution is within a predetermined range; performing spray drying on the concentrated solution to obtain evodia rutaecarpa spray-dried powder; and granulating the evodia rutaecarpa spray-dried powder to obtain evodia rutaecarpa formula granules. According to the present invention, in combination with an artificial intelligence technology, whole-process monitoring, tracking and tracing of traditional Chinese medicinal materials and online detection, evaluation and feedback of traditional Chinese medicine formula granules are realized, thereby improving the quality safety and traceability.

Description

Preparation method and system of Evodia rutaecarpa formula granule preparation Technical Field

[0001] The present invention relates to the field of intelligent preparation technology, and in particular to a preparation method and a system thereof for a granular preparation of Evodia rutaecarpa. Background Art

[0002] Chinese herbal formula granules are made by processing Chinese herbal medicine slices through various processes under the guidance of traditional Chinese medicine theory. They are granules used in the preparation of clinical Chinese medicine prescriptions. Compared with traditional Chinese herbal decoctions, they are easy to take, require a small dosage, have controllable quality, are easy to store, and are convenient to carry.

[0003] Evodia rutaecarpa is a commonly used Chinese medicinal herb with warming, cold-relieving, analgesic, and insecticidal properties. It is primarily used to treat stomach cold, vomiting, abdominal pain, and ascariasis. However, the preparation method of Evodia rutaecarpa granules significantly impacts their stability, dissolution, and bioavailability. Current methods for preparing Evodia rutaecarpa granules exhibit significant discrepancies in the active ingredient content and chemical fingerprints of the raw materials, impacting the stability of the therapeutic efficacy of the granules.

[0004] Therefore, an optimized preparation scheme for the Evodia rutaecarpa formula granule preparation is expected.

[0005] Summary of the Invention

[0006] The embodiment of the present invention provides a method and system for preparing a formula granule preparation of Evodia rutaecarpa, comprising the steps of: firstly boiling Evodia rutaecarpa slices with water to obtain a first decoction liquid and medicinal residues; secondly boiling the medicinal residues with water to obtain a second decoction liquid; mixing the first decoction liquid and the second decoction liquid and concentrating them under reduced pressure to obtain a concentrated solution, wherein the relative density of the concentrated solution is within a predetermined range; spray drying the concentrated solution to obtain Evodia rutaecarpa spray-dried powder; and granulating the Evodia rutaecarpa spray-dried powder to obtain Evodia rutaecarpa formula granules. In this way, the method can be combined with artificial intelligence technology to achieve full-process monitoring, tracking, and traceability of Chinese medicinal materials, as well as online detection, evaluation, and feedback of Chinese medicinal formula granules, thereby improving quality safety and traceability.

[0007] The present invention also provides a method for preparing a granular preparation of Evodia rutaecarpa, which comprises:

[0008] The Evodia rutaecarpa slices are first decocted with water to obtain a decoction and medicinal residues;

[0009] Decoctioning the medicinal residue for the second time to obtain a second decoction;

[0010] mixing the first decoction liquid and the second decoction liquid and concentrating them under reduced pressure to obtain a concentrated solution, wherein the relative density of the concentrated solution is within a predetermined range;

[0011] spray-drying the concentrated solution to obtain Evodia rutaecarpa spray-dried powder; and

[0012] The spray-dried Evodia rutaecarpa powder is granulated to obtain Evodia rutaecarpa formula granules.

[0013] The embodiment of the present invention further provides a system for preparing a granular preparation of Evodia rutaecarpa, which comprises:

[0014] The first decoction module is used to add water and decoct the Evodia rutaecarpa slices for the first time to obtain a decoction liquid and medicinal residues;

[0015] a second decoction module, for decoct the medicinal residue for the second time to obtain a second decoction liquid;

[0016] a decompression and concentration module, configured to mix the first decoction liquid and the second decoction liquid and concentrate them under reduced pressure to obtain a concentrated liquid, wherein the relative density of the concentrated liquid is within a predetermined range;

[0017] a spray drying module, for spray drying the concentrated liquid to obtain Evodia rutaecarpa spray-dried powder; and

[0018] The granulation module is used for granulating the spray-dried Evodia rutaecarpa powder to obtain Evodia rutaecarpa formula granules. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0020] FIG1 is a flow chart of a method for preparing a granular preparation of Evodia rutaecarpa provided in an embodiment of the present invention.

[0021] FIG2 is a schematic diagram of the system architecture of a method for preparing an Evodia rutaecarpa formula granule preparation provided in an embodiment of the present invention.

[0022] FIG3 is a flow chart of the sub-steps of step 130 in the method for preparing an Evodia rutaecarpa formula granule preparation provided in an embodiment of the present invention.

[0023] FIG4 is a block diagram of a system for preparing an Evodia rutaecarpa granule preparation provided in an embodiment of the present invention.

[0024] FIG5 is an application scenario diagram of a method for preparing an Evodia rutaecarpa formula granule preparation provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0025] To make the purpose, technical solutions and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below with reference to the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0026] Unless otherwise specified, all technical and scientific terms used in the embodiments of the present application have the same meaning as those commonly understood by those skilled in the art in the art of the present application. The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit the scope of this application.

[0027] In the description of the embodiments of this application, it should be noted that, unless otherwise specified and limited, the term "connection" should be understood in a broad sense. For example, it can be an electrical connection, or it can be the internal connection between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meaning of the above terms can be understood according to the specific circumstances.

[0028] It should be noted that the terms "first, second, and third" in the embodiments of the present application are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It is understood that the terms "first, second, and third" can be interchanged to represent a specific order or precedence where permitted. It should be understood that the objects distinguished by "first, second, and third" can be interchanged where appropriate, such that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0029] Evodia rutaecarpa, also known as Chinese gallnut, is a common Chinese medicinal herb belonging to the Rhamnaceae family. It primarily grows in southern China, such as Jiangsu, Zhejiang, Fujian, and Guangdong, with smaller quantities also found in Japan and North Korea.

[0030] The fruit of Evodia rutaecarpa is conical, red or purple when ripe, and contains seeds. Its pungent, warm, and slightly cooling properties promote qi circulation, dispel cold, and relieve vomiting. It is widely used in traditional Chinese medicine formulas, particularly for treating gastrointestinal disorders, vomiting, and chest tightness.

[0031] Evodia rutaecarpa contains a variety of active ingredients, including evodiol, evodiac acid, and strychnine. These ingredients have anti-inflammatory, antioxidant, antibacterial, and anti-tumor effects, and are effective in improving digestion, promoting blood circulation, and relieving pain.

[0032] In addition to its medicinal value, Evodia rutaecarpa is also widely used in food and condiments. Its fruit can be used to make sauces, seasonings, teas, etc. In addition, Evodia rutaecarpa also has certain ornamental value. Its red fruit gives people a unique visual enjoyment in autumn.

[0033] It should be noted that although Evodia rutaecarpa has certain medicinal value, it should be used under the guidance of a doctor or Chinese medicine practitioner to avoid excessive use or adverse reactions with other drugs.

[0034] In one embodiment of the present invention, FIG1 is a flow chart of a method for preparing a formula granule preparation of Evodia rutaecarpa provided in an embodiment of the present invention. As shown in FIG1 , a method 100 for preparing a formula granule preparation of Evodia rutaecarpa according to an embodiment of the present invention includes: 110, decocting Evodia rutaecarpa slices for a first time to obtain a first decoction and medicinal residues; 120, decocting the medicinal residues for a second time to obtain a second decoction; 130, mixing the first decoction and the second decoction and concentrating them under reduced pressure to obtain a concentrated solution, wherein the relative density of the concentrated solution is within a predetermined range; 140, spray-drying the concentrated solution to obtain a spray-dried Evodia rutaecarpa powder; and 150, granulating the spray-dried Evodia rutaecarpa powder to obtain formula granules of Evodia rutaecarpa.

[0035] In a specific embodiment of the present application, the Evodia rutaecarpa formula granules are prepared:

[0036] 100 kg of processed Evodia rutaecarpa slices are taken, 800 kg of water is added, and the mixture is extracted for 2 hours, i.e., the first decoction; 800 kg of water is added to the residue after the first decoction, and the mixture is extracted for 1.5 hours, i.e., the second decoction; the two decoctions are combined and concentrated under reduced pressure to a relative density of 1.05-1.10; the concentrated solution is preheated at 45° C., an air inlet temperature of 180° C., a material pump speed of 550 rpm, an air outlet temperature of 80° C., and an air delivery temperature of 45° C. to obtain processed Evodia rutaecarpa spray-dried powder; 0-5% by weight of the auxiliary material dextrin is added to the processed Evodia rutaecarpa spray-dried powder; the processed Evodia rutaecarpa spray-dried powder is dry-granulated under the conditions of a main pressure wheel pressure of 5 MPa, a side pressure wheel pressure of 0.5 MPa, a main shaft speed of 400 rpm, and a feeding voltage of 120 V to obtain 20-60 mesh processed Evodia rutaecarpa formula granules, wherein each gram of the processed Evodia rutaecarpa formula granules is equivalent to 5 grams of the raw medicinal material.

[0037] In another specific embodiment of the present application, the Evodia rutaecarpa formula granules are prepared:

[0038] 100 kg of processed Evodia rutaecarpa slices were taken, 1000 kg of water was added, and the mixture was extracted for 2 hours, i.e., the first decoction; 600 kg of water was added to the residue after the first decoction, and the mixture was extracted for 1 hour, i.e., the second decoction; the two decoctions were combined and concentrated under reduced pressure to a relative density of 1.00-1.05; the concentrate was preheated at 50° C., an air inlet temperature of 170° C., a material pump speed of 600 rpm, an air outlet temperature of 75° C., and an air delivery temperature of 50° C. to obtain a processed Evodia rutaecarpa spray-dried powder; 5-15% by weight of the auxiliary material dextrin was added to the processed Evodia rutaecarpa spray-dried powder; dry granulation was performed under the conditions of a main pressing wheel pressure of 6 MPa, a side pressing wheel pressure of 0.4 MPa, a main shaft speed of 450 rpm, and a feeding voltage of 110 V to obtain 20-60 mesh processed Evodia rutaecarpa formula granules, wherein each gram of the processed Evodia rutaecarpa formula granules was equivalent to 4 grams of the raw medicinal material;

[0039] The quality control method of the prepared Evodia rutaecarpa slices comprises the following steps: using infrared fingerprint spectrum to identify the quality of the prepared Evodia rutaecarpa slices, and the infrared fingerprint spectrum identification method is: using a Fourier transform infrared spectrometer with a measurement range of 4000 cm -1 -400cm -1 , DTGS detector, resolution 4cm -1 , scanning times 16 times, water and carbon dioxide interference were deducted from the scanning process, the relative humidity of the environment was lower than 60%, and the potassium bromide direct compression method was used for detection; the infrared fingerprint of the prepared Evodia rutaecarpa slices is shown in Figure 2, with the following characteristics: at 1614 cm -1 、1444cm -1 、1262cm -1 、1063cm -1 There is obvious absorption at 1614cm -1 、1063cm -1 The absorption peak intensity at 2922cm -1 、763cm -1 There is characteristic absorption.

[0040] In another specific embodiment of the present application, the Evodia rutaecarpa formula granules are prepared:

[0041] 100 kg of processed Evodia rutaecarpa slices were taken, 1400 kg of water was added, and the mixture was extracted for 2.5 hours, i.e., the first decoction; 1100 kg of water was added to the residue after the first decoction, and the mixture was extracted for 2.5 hours, i.e., the second decoction; the two decoctions were combined and concentrated under reduced pressure to a relative density of 1.10-1.15; the concentrate was preheated at 45° C., an inlet air temperature of 170° C., a feed pump speed of 600 rpm, an outlet air temperature of 75° C., and a wind supply temperature of 50° C. to obtain processed Evodia rutaecarpa spray-dried powder; 15-25% by weight of lactose was added to the processed Evodia rutaecarpa spray-dried powder; dry granulation was performed under the conditions of a main pressing wheel pressure of 4 MPa, a side pressing wheel pressure of 0.4 MPa, a main shaft speed of 500 rpm, and a feeding voltage of 130 V to obtain 20-60 mesh processed Evodia rutaecarpa formula granules, wherein each gram of the processed Evodia rutaecarpa formula granules was equivalent to 3 grams of the crude drug;

[0042] The quality control method of the obtained Evodia rutaecarpa granules comprises the following steps: using infrared fingerprint spectrum for quality identification, the infrared fingerprint spectrum identification method is: using Fourier transform infrared spectrometer, the measurement range is 4000cm -1 -400cm -1 , DTGS detector, resolution 4cm -1 , scanning times 16 times, water and carbon dioxide interference were deducted from the scanning process, the relative humidity of the environment was lower than 60%, and the potassium bromide direct tableting method was used for detection; the infrared fingerprint of the prepared Evodia rutaecarpa formula granules is shown in Figure 1, with the following characteristics: at 1609cm -1 There is a maximum absorption peak at 2929cm -1 、1404cm -1 、1284cm-1、1074cm -1 、817cm -1 There is characteristic absorption.

[0043] In step 130, it is usually necessary to manually measure the relative density of the concentrated liquid to ensure that the relative density of the concentrated liquid is within a predetermined range. However, this process requires a lot of time and manpower costs, resulting in low production efficiency.

[0044] Artificial intelligence technology is a computer system that simulates human intelligence, capable of performing complex tasks such as learning, reasoning, and judgment. With the gradual development of AI technology, it has also seen widespread application in the preparation of traditional Chinese medicine granules. For example, AI technology can improve the efficiency and quality of traditional Chinese medicine granule preparation. Through methods such as image recognition, speech recognition, and natural language processing, it enables automated identification, sorting, and weighing of Chinese medicinal materials, as well as automated packaging and labeling of these granules. This reduces human error and contamination, ensuring the standardization of the preparation process. Furthermore, AI technology can enhance the quality control and traceability of traditional Chinese medicine granules. Through sensor networks, the Internet of Things, and blockchain, it enables full-process monitoring, tracking, and tracing of Chinese medicinal materials, as well as online testing, evaluation, and feedback for these granules, improving quality, safety, and traceability. The development and application of AI technology in the preparation of traditional Chinese medicine granules will help promote the modernization and internationalization of traditional Chinese medicine and enhance the competitiveness and influence of the traditional Chinese medicine industry.

[0045] In this regard, in response to the above technical problems, the technical concept of the present application is to automatically determine whether the relative density of the concentrated liquid is within a predetermined range through image detection combined with artificial intelligence technology.

[0046] It's understandable that traditionally, the relative density of concentrates requires laboratory testing, which consumes time and human resources. However, the use of image detection and artificial intelligence technologies can automate detection and judgment, significantly improving production efficiency. Manual relative density measurement is subject to subjectivity and error, but the use of image detection and artificial intelligence technologies can reduce human interference and improve measurement accuracy and reliability.

[0047] Furthermore, relative density is a key indicator in the concentrate production process, directly impacting product quality. Automated detection and assessment of relative density allows for timely detection of anomalies and the implementation of appropriate measures, thereby improving quality control. Leveraging image detection and artificial intelligence technologies, the relative density of concentrates can be monitored in real time, providing timely feedback. This facilitates timely adjustments to the production process to ensure that products meet predetermined quality standards.

[0048] The use of image detection and artificial intelligence technology to automatically determine whether the relative density of the concentrate is within a predetermined range is necessary to improve production efficiency, enhance quality control, reduce human errors, and provide real-time monitoring feedback.

[0049] FIG2 is a schematic diagram of the system architecture of a method for preparing a granule preparation of Evodia rutaecarpa provided in an embodiment of the present invention. FIG3 is a flowchart of the sub-steps of step 130 in a method for preparing a granule preparation of Evodia rutaecarpa provided in an embodiment of the present invention. As shown in FIG2 and FIG3 , the first decoction liquid and the second decoction liquid are mixed and concentrated under reduced pressure to obtain a concentrated liquid, wherein the relative density of the concentrated liquid is within a predetermined range, comprising: 131, acquiring a detection image of the concentrated liquid captured by a camera; 132, extracting concentrated liquid feature information from the detection image of the concentrated liquid to obtain a deep and shallow concentrated liquid texture feature matrix; and, 133, determining whether the relative density of the concentrated liquid is within a predetermined range based on the deep and shallow concentrated liquid texture feature matrix.

[0050] In step 131, the position and angle of the camera are ensured to be appropriate to obtain a clear image of the concentrated liquid. The concentrated liquid image is captured by the camera to provide a data basis for subsequent image processing and analysis.

[0051] In step 132, image processing techniques, such as image segmentation and feature extraction, are used to extract feature information related to the dark and light textures of the concentrated liquid from the concentrated liquid image. By extracting this feature information from the concentrated liquid image, a texture feature matrix related to the relative density of the concentrated liquid can be obtained, providing a basis for subsequent relative density determination.

[0052] In step 133, a machine learning or deep learning algorithm is used to train a model based on the texture feature matrix of the dark and light concentrates to determine whether the relative density of the concentrate is within a predetermined range. This determination based on the texture feature matrix of the dark and light concentrates can automatically determine whether the relative density of the concentrate is within the predetermined range, thereby improving production efficiency and quality control.

[0053] Automating the relative density determination of concentrates through image detection and artificial intelligence technology can reduce human error and improve production efficiency and quality control. Furthermore, using a matrix of texture features of dark and light concentrates to determine relative density improves accuracy and reliability, providing real-time monitoring and feedback for the production process.

[0054] Specifically, in step 131, a detection image of the concentrated liquid captured by the camera is obtained. In the technical solution of the present application, first, a detection image of the concentrated liquid captured by the camera is obtained. The detection image of the concentrated liquid captured by the camera can obtain: 1. The color of the concentrated liquid. The color of the concentrated liquid can provide some clues about its concentration. Generally speaking, the darker the color, the higher the concentration. 2. The transparency of the concentrated liquid. The transparency can reflect the concentration of the solute in the concentrated liquid. If the concentrated liquid is relatively turbid, it may mean that the concentration is high. 3. The liquid level of the concentrated liquid. The liquid level of the concentrated liquid can provide information about the volume. If the liquid level is low, it may mean that the volume of the concentrated liquid is small. The foaming condition of the concentrated liquid. The presence of foam can indicate the gas content in the concentrated liquid, and the presence of gas may affect the relative density of the concentrated liquid.

[0055] By analyzing this image information and combining it with artificial intelligence technology, it is possible to determine whether the relative density of the concentrate is within a predetermined range, thereby achieving automated quality control. In this application, the detection images captured by the camera provide intuitive visual information about the appearance of the concentrate. These images can show the color, texture, transparency and other characteristics of the concentrate, and these characteristics are related to the relative density. Through the images captured by the camera, visual data of the concentrate can be obtained, providing a data basis for subsequent image processing and analysis. Based on this data, characteristic information of the concentrate can be extracted for relative density judgment. From the images captured by the camera, image processing technology can be used to extract characteristic information related to the relative density of the concentrate. For example, texture features, color features, etc. in the image can be extracted, and these features are related to the relative density of the concentrate.

[0056] The detection image of the concentrated liquid captured by the camera provides visual information and data acquisition basis for the final determination of whether the relative density of the concentrated liquid is within the predetermined range, as well as basic data for feature extraction, model training and judgment, thereby playing an important role in the process of automatically judging the relative density of the concentrated liquid.

[0057] Specifically, in step 132, the concentrated liquid feature information in the detection image of the concentrated liquid is extracted to obtain a deep and shallow concentrated liquid texture feature matrix, including: extracting the texture feature information of the detection image of the concentrated liquid to obtain a concentrated liquid texture feature matrix; and improving the feature expression ability of the concentrated liquid texture feature matrix to obtain a deep and shallow concentrated liquid texture feature matrix.

[0058] The method of extracting texture feature information of the detection image of the concentrated liquid to obtain a concentrated liquid texture feature matrix includes: calculating the gradient histogram of the detection image of the concentrated liquid to obtain a concentrated liquid gradient histogram; and passing the concentrated liquid gradient histogram through an image texture feature extractor based on a convolutional neural network model including a spatial attention module to obtain the concentrated liquid texture feature matrix.

[0059] Next, texture feature information is extracted from the detected image of the concentrated liquid to obtain a concentrated liquid texture feature matrix. Specifically, the spatial, textural, and structural feature information contained in the detected image of the concentrated liquid is extracted. Specifically, spatial feature information can include the spatial distribution and shape information of different components in the concentrated liquid, thereby revealing the relative position, size, and shape of the components. Texture feature information can provide information about the surface characteristics of the concentrated liquid, including lines, texture direction, texture density, etc. Structural feature information can reveal the structural and organizational characteristics of the concentrated liquid. For example, it can reflect information such as the hierarchical structure of the concentrated liquid.

[0060] In a specific example of the present application, the encoding process of extracting the texture feature information of the detection image of the concentrated liquid to obtain the concentrated liquid texture feature matrix includes: first calculating the gradient histogram of the detection image of the concentrated liquid to obtain a concentrated liquid gradient histogram; then, passing the concentrated liquid gradient histogram through an image texture feature extractor based on a convolutional neural network model including a spatial attention module to obtain a concentrated liquid texture feature matrix.

[0061] The gradient histogram is a method used to describe image texture features, reflecting the variation in pixel grayscale values ​​within an image. A gradient histogram is constructed by calculating the gradient value (i.e., the rate of change of the pixel grayscale value) for each pixel in the image. The gradient refers to the direction and magnitude of the grayscale change at each pixel in an image. When calculating a gradient histogram, the Sobel operator or other gradient operators are typically used to calculate the gradient value for each pixel in the image. The gradient values ​​are then divided into a certain number of intervals (usually 16 or 32 intervals), and the number of pixels in each interval is counted to obtain the gradient histogram.

[0062] The gradient histogram can reflect image details such as edges and textures. For concentrated liquid detection images, calculating the gradient histogram can extract the concentrated liquid's texture features, such as particles and liquid level changes. By analyzing the shape and distribution of the gradient histogram, the relative density of the concentrated liquid can be preliminarily determined and classified.

[0063] It should be noted that the gradient histogram is only one method of representing texture features, and it can also be combined with other feature extraction methods for comprehensive analysis to obtain more comprehensive texture feature information of the concentrate.

[0064] Computing the gradient histogram of the concentrated liquid test image can extract the concentrated liquid's texture features, thereby facilitating its classification and analysis. The gradient histogram can capture texture variations in the concentrated liquid image, such as edges, textures, and details. By calculating the gradient histogram, these texture features can be extracted to describe the concentrated liquid's texture characteristics.

[0065] The gradient histogram of a concentrated solution can reflect the concentration changes of particles or solutes in the concentrated solution. By comparing the gradient histograms at different concentrations, the differences between concentrated solutions of different concentrations can be identified, thereby enabling concentration classification and differentiation. The gradient histogram can help detect anomalies in concentrated solution images, such as uneven particle distribution and abnormal solute concentration. By comparing the gradient histogram with the gradient histogram of a normal concentrated solution, anomalies can be identified and addressed promptly.

[0066] Gradient histograms can be used as indicators for concentrate quality control. By establishing a reference model for the concentrate gradient histogram, actual concentrate gradient histograms can be compared to determine whether the concentrate meets quality requirements. Calculating the concentrate gradient histogram can extract texture features, helping to distinguish different concentrations and detect anomalies. This can be used for concentrate quality control, thereby improving production efficiency and product quality.

[0067] Furthermore, spatial attention is a technique used to enhance the convolutional neural network model's focus on spatial information in images. Traditional convolutional neural networks typically focus only on local features at each location when processing images, without fully considering the relationships between different locations. Spatial attention introduces learnable weight parameters, allowing the model to automatically learn the correlations between different locations and adjust feature weights based on these correlations. This allows the model to pay more attention to important spatial locations when extracting features, thereby improving the model's ability to perceive important information in the image.

[0068] There are various ways to implement spatial attention in convolutional neural networks. One common approach is to use the self-attention mechanism. The self-attention mechanism calculates the similarity between different locations in the input feature map and then adjusts the feature weights based on this similarity. This allows the model to automatically learn the correlations between different locations based on the specific content of the input image.

[0069] By introducing spatial attention, convolutional neural networks can better capture spatial relationships in images, thereby improving model performance. In the task of extracting image texture features, spatial attention can help the model better focus on texture details in the gradient histogram of the concentrated solution, improving the accuracy and expressiveness of feature extraction.

[0070] Furthermore, the feature expression capability of the concentrated liquid texture feature matrix is ​​improved to obtain the deep and shallow concentrated liquid texture feature matrix, including: using the residual idea to fuse the concentrated liquid gradient histogram and the concentrated liquid texture feature matrix to obtain the deep and shallow concentrated liquid texture feature matrix.

[0071] Then, the residual concept is used to fuse the concentrate gradient histogram and the concentrate texture feature matrix to obtain a deep and shallow concentrate texture feature matrix. This method improves the feature expression capability of the concentrate texture feature matrix so that the deep and shallow concentrate texture feature matrix contains both the relatively shallow information expressed by the concentrate gradient histogram, such as the concentration changes of the concentrate components, and the deeper information expressed by the concentrate texture feature matrix, such as the spatial distribution, texture characteristics, and potential structure of the concentrate components.

[0072] Residual learning is a key concept in deep learning, used to address the vanishing and exploding gradient problems during deep neural network training. This concept was first proposed by Kaiming He et al. in 2015 and has been widely used in Residual Network (ResNet).

[0073] In traditional deep neural networks, the network is constructed by stacking multiple layers, each of which transforms the input using a nonlinear activation function. However, as the number of network layers increases, the gradient tends to decay or explode during backpropagation, making the network difficult to train. The residual concept addresses this problem by introducing the "residual block." The residual block contains a skip connection that adds the input directly to the output, allowing the network to learn the residual function. Specifically, the output of the residual block can be expressed as: output = input + residual function (input).

[0074] This design makes it easier for the network to learn the identity mapping, that is, the mapping that directly transfers the input to the output. If the residual function is close to zero, the network can choose to ignore the function, thereby achieving direct transfer between layers. The use of residual thinking can effectively solve the problems of gradient vanishing and gradient exploding, making the training of deep networks more stable and efficient. In the process of fusing the concentrated liquid gradient histogram and the concentrated liquid texture feature matrix, the residual idea can be borrowed to obtain the deep and shallow concentrated liquid texture feature matrix by adding or subtracting the features of the two, thereby better representing the characteristic information of the concentrated liquid.

[0075] Furthermore, the concentrate gradient histogram and concentrate texture feature matrix each capture distinct characteristics of the concentrate. Fusion can compensate for missing information in each, providing a more comprehensive representation of the features. The residual concept allows the network to learn a residual mapping, meaning it can learn incremental information about the concentrate's features. Through residual connections, the fused features can better represent the texture characteristics of both shallow and deep concentrates, improving feature representation. Residual connections allow gradients to be directly propagated to shallower layers, alleviating the vanishing gradient problem. This is crucial for training deep networks, as it allows for more efficient gradient propagation and facilitates network convergence.

[0076] Through the above results, fusing the concentrate gradient histogram and the concentrate texture feature matrix can produce a more powerful feature representation, improving the accuracy and robustness of the texture features of deep and shallow concentrates. This has a positive impact on quality control and production efficiency in the preparation of traditional Chinese medicine granules.

[0077] Specifically, in step 133, based on the deep and shallow concentrated liquid texture feature matrix, determining whether the relative density of the concentrated liquid is within a predetermined range includes: passing the deep and shallow concentrated liquid texture feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the relative density of the concentrated liquid is within a predetermined range.

[0078] Then, the texture feature matrix of the deep and shallow concentrated liquid is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the relative density of the concentrated liquid is within a predetermined range.

[0079] Furthermore, in the present application, the preparation method of the Evodia rutaecarpa formula granule preparation further includes a training step: training the image texture feature extractor and the classifier based on the convolutional neural network model including the spatial attention module; wherein the training step includes: obtaining training data, the training data including a training detection image of the concentrate, and whether the relative density of the concentrate is within a predetermined range; calculating the gradient histogram of the training detection image of the concentrate to obtain a training concentrate gradient histogram; and subjecting the training concentrate gradient histogram to the image texture feature extractor based on the convolutional neural network model including the spatial attention module. to obtain a training concentrate texture feature matrix; to use the residual idea to fuse the training concentrate gradient histogram and the training concentrate texture feature matrix to obtain a training deep and shallow concentrate texture feature matrix; to pass the training deep and shallow concentrate texture feature matrix through a classifier to obtain a classification loss function value; to calculate the shared manifold implicit similarity factor of the concentrate gradient histogram and the concentrate texture feature matrix; and to train the image texture feature extractor based on the convolutional neural network model including the spatial attention module and the classifier using the weighted sum of the classification loss function value and the shared manifold implicit similarity factor as the loss function value.

[0080] Calculating the shared manifold implicit similarity factor of the concentrated liquid gradient histogram and the concentrated liquid texture feature matrix includes: expanding the concentrated liquid gradient histogram and the concentrated liquid texture feature matrix into concentrated liquid gradient orthogonal vectors and concentrated liquid texture feature vectors; and calculating the shared manifold implicit similarity factor of the concentrated liquid gradient orthogonal vectors and the concentrated liquid texture feature vectors.

[0081] In the technical solution of the present application, when the concentrated liquid gradient histogram passes through an image texture feature extractor based on a convolutional neural network model including a spatial attention module, the concentrated liquid texture feature matrix is ​​used to express the image semantic features enhanced by the local spatial distribution of the concentrated liquid gradient histogram. Therefore, the concentrated liquid texture feature matrix has differences in both the semantic source-feature mode and the semantic space distribution mode relative to the concentrated liquid gradient histogram as the source image. In this way, when the concentrated liquid gradient histogram and the concentrated liquid texture feature matrix are fused through the residual idea, the high-dimensional feature manifold of the deep and shallow concentrated liquid texture feature matrix as a feature representation in a high-dimensional feature space will have a problem of poor geometric monotonicity, which affects its convergence effect when performing classification regression through a classifier, thereby affecting the accuracy of the classification results.

[0082] Therefore, in order to solve the problem of differences between the concentrated liquid gradient histogram and the concentrated liquid texture feature matrix in semantic source-feature mode and semantic space distribution mode, the applicant of this application first expands the concentrated liquid gradient histogram and the concentrated liquid texture feature matrix into concentrated liquid gradient vertical vectors and concentrated liquid texture feature vectors to explicitly arrange their semantic space distributions, and then introduces a common manifold implicit similarity factor as a loss function for the concentrated liquid gradient vertical vector V1 and the concentrated liquid texture feature vector V2, which is specifically expressed as:

[0083] Among them, ||·||2 represents the two-norm of the vector, and represents the square root of the Frobenius norm of the matrix, V1 is the vertical vector of the concentrated liquid gradient, V2 is the texture feature vector of the concentrated liquid, V2 T is the transposed vector of the concentrated liquid texture feature vector, the concentrated liquid gradient vertical vector and the concentrated liquid texture feature vector are both in column vector form, w1, w2, w3 and α are weight hyperparameters, Represents vector multiplication, ⊙ represents point multiplication by position, represents the difference by position, represents the implicit similarity factor loss function value of the common manifold.

[0084] Here, the implicit similarity factor of the shared manifold can be used to represent the shared manifold of each feature manifold in the cross dimension by the structured association between the concentrated liquid gradient vertical vector V1 and the concentrated liquid texture feature vector V2, and share the common constraints of manifold structural factors such as the difference, correspondence, and correlation of the concentrated liquid gradient vertical vector V1 and the concentrated liquid texture feature vector V2 with the same factorized weight, thereby measuring the distribution similarity of the geometric derivative structure representation that depends on the shared manifold, so as to achieve the nonlinear geometric monotonicity of the residual fusion features between cross-modal features, and improve the convergence effect of the deep and shallow concentrated liquid texture feature matrix when performing classification regression through the classifier, thereby improving the accuracy of the classification results.

[0085] In summary, the preparation method 100 of the Evodia rutaecarpa formula granule preparation according to an embodiment of the present invention is explained, which uses image detection combined with artificial intelligence technology to automatically determine whether the relative density of the concentrated solution is within a predetermined range.

[0086] FIG4 is a block diagram of a system for preparing an Evodia rutaecarpa granule preparation according to an embodiment of the present invention. As shown in FIG4 , the system for preparing the Evodia rutaecarpa granule preparation includes: a first decoction module 210 for performing a first decoction of Evodia rutaecarpa slices with water to obtain a first decoction liquid and medicinal residues; a second decoction module 220 for performing a second decoction of the medicinal residue with water to obtain a second decoction liquid; a vacuum concentration module 230 for mixing the first and second decoction liquids and concentrating them under reduced pressure to obtain a concentrated solution, wherein the relative density of the concentrated solution is within a predetermined range; a spray drying module 240 for spray drying the concentrated solution to obtain an Evodia rutaecarpa spray-dried powder; and a granulation module 250 for granulating the spray-dried Evodia rutaecarpa powder to obtain the Evodia rutaecarpa granules.

[0087] Those skilled in the art will appreciate that the specific operations of each step in the preparation system of the above-mentioned Evodia rutaecarpa formula granule preparation have been described in detail in the description of the preparation method of the Evodia rutaecarpa formula granule preparation with reference to Figures 1 to 3 above, and therefore, repeated description thereof will be omitted.

[0088] As described above, the system 100 for preparing an Evodia rutaecarpa granule preparation according to an embodiment of the present invention can be implemented in various terminal devices, such as a server for preparing an Evodia rutaecarpa granule preparation. In one example, the system 100 for preparing an Evodia rutaecarpa granule preparation according to an embodiment of the present invention can be integrated into a terminal device as a software module and / or a hardware module. For example, the system 100 for preparing an Evodia rutaecarpa granule preparation can be a software module in the operating system of the terminal device, or can be an application developed specifically for the terminal device; of course, the system 100 for preparing an Evodia rutaecarpa granule preparation can also be one of the many hardware modules of the terminal device.

[0089] Alternatively, in another example, the preparation system 100 of the Evodia rutaecarpa formula granule preparation and the terminal device may also be separate devices, and the preparation system 100 of the Evodia rutaecarpa formula granule preparation may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0090] Figure 5 is an application scenario diagram of a method for preparing an Evodia rutaecarpa granule preparation provided in an embodiment of the present invention. As shown in Figure 5 , in this application scenario, first, a detection image of the concentrated liquid captured by a camera is acquired (e.g., as shown in C in Figure 5 ); then, the acquired detection image of the concentrated liquid is input into a server (e.g., as shown in S in Figure 5 ) that is equipped with a preparation algorithm for the Evodia rutaecarpa granule preparation. The server is capable of processing the detection image of the concentrated liquid based on the preparation algorithm for the Evodia rutaecarpa granule preparation to determine whether the relative density of the concentrated liquid is within a predetermined range.

[0091] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for preparing a granular preparation of Evodia rutaecarpa, characterized in that: include: The Evodia rutaecarpa decoction pieces are firstly decocted with water to obtain a decoction liquid and a medicinal residue; Decoction the medicinal residue for the second time to obtain a second decoction liquid; The first decoction liquid and the second decoction liquid are mixed and concentrated under reduced pressure to obtain a concentrated solution, wherein the relative density of the concentrated solution is within a predetermined range; spray-drying the concentrated solution to obtain Evodia rutaecarpa spray-dried powder; and The spray-dried powder of Evodia rutaecarpa is granulated to obtain the Evodia rutaecarpa formula granules.

2. The method for preparing the Evodia rutaecarpa granule preparation according to claim 1, characterized in that: The first decoction liquid and the second decoction liquid are mixed and concentrated under reduced pressure to obtain a concentrated solution, wherein the relative density of the concentrated solution is within a predetermined range, comprising: Acquiring a detection image of the concentrated liquid captured by a camera; Extracting concentrated liquid feature information from the detection image of the concentrated liquid to obtain a deep and shallow concentrated liquid texture feature matrix; and Based on the deep and shallow concentrated liquid texture feature matrix, it is determined whether the relative density of the concentrated liquid is within a predetermined range.

3. The method for preparing the Evodia rutaecarpa granule preparation according to claim 2, characterized in that: Extracting concentrated liquid feature information from the detection image of the concentrated liquid to obtain a deep and shallow concentrated liquid texture feature matrix includes: Extracting texture feature information of the detection image of the concentrated liquid to obtain a concentrated liquid texture feature matrix; and The feature expression capability of the concentrated liquid texture feature matrix is ​​improved to obtain a deep and shallow concentrated liquid texture feature matrix.

4. The method for preparing the Evodia rutaecarpa granule preparation according to claim 3, characterized in that: Extracting texture feature information of the detection image of the concentrated liquid to obtain a concentrated liquid texture feature matrix includes: Calculating a gradient histogram of the detection image of the concentrated liquid to obtain a concentrated liquid gradient histogram; and The concentrated liquid gradient histogram is passed through an image texture feature extractor based on a convolutional neural network model including a spatial attention module to obtain the concentrated liquid texture feature matrix.

5. The method for preparing the Evodia rutaecarpa granule preparation according to claim 4, characterized in that: Improving the feature expression capability of the concentrated liquid texture feature matrix to obtain a deep and shallow concentrated liquid texture feature matrix includes: The residual idea is used to fuse the concentrated liquid gradient histogram and the concentrated liquid texture feature matrix to obtain the deep and shallow concentrated liquid texture feature matrix.

6. The method for preparing the Evodia rutaecarpa granule preparation according to claim 5, characterized in that: Determining whether the relative density of the concentrated liquid is within a predetermined range based on the deep and shallow concentrated liquid texture feature matrix includes: The texture feature matrix of the deep and shallow concentrated liquid is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the relative density of the concentrated liquid is within a predetermined range.

7. The method for preparing the Evodia rutaecarpa granule preparation according to claim 6, characterized in that: The method further includes a training step of: training the image texture feature extractor based on the convolutional neural network model including the spatial attention module and the classifier; Wherein, the training step includes: Acquiring training data, the training data including a training detection image of the concentrated liquid and whether the relative density of the concentrated liquid is within a predetermined range; Calculating a gradient histogram of the training detection image of the concentrated liquid to obtain a training concentrated liquid gradient histogram; Passing the training concentrate gradient histogram through the image texture feature extractor based on the convolutional neural network model including the spatial attention module to obtain a training concentrate texture feature matrix; Using the residual idea to fuse the training concentrated liquid gradient histogram and the training concentrated liquid texture feature matrix to obtain the training deep and shallow concentrated liquid texture feature matrix; Passing the training depth and shallowness concentrated liquid texture feature matrix through a classifier to obtain a classification loss function value; Calculating a common manifold implicit similarity factor of the concentrated liquid gradient histogram and the concentrated liquid texture feature matrix; and The image texture feature extractor based on the convolutional neural network model including the spatial attention module and the classifier are trained using the weighted sum of the classification loss function value and the common manifold implicit similarity factor as the loss function value.

8. The method for preparing the Evodia rutaecarpa granule preparation according to claim 7, characterized in that: Calculating the common manifold implicit similarity factor of the concentrated liquid gradient histogram and the concentrated liquid texture feature matrix, comprising: Expanding the concentrated liquid gradient histogram and the concentrated liquid texture feature matrix into concentrated liquid gradient histogram vectors and concentrated liquid texture feature vectors; and The common manifold implicit similarity factor of the concentrated liquid gradient vector and the concentrated liquid texture feature vector is calculated.

9. The method for preparing the Evodia rutaecarpa granule preparation according to claim 8, characterized in that: Calculating the common manifold implicit similarity factor of the concentrated liquid gradient vertical vector and the concentrated liquid texture feature vector, comprising: calculating the common manifold implicit similarity factor of the concentrated liquid gradient vertical vector and the concentrated liquid texture feature vector using the following optimization formula as a loss function; Wherein, the optimization formula is: Among them, ||·||2 represents the bi-norm of the vector, and represents the square root of the Frobenius norm of the matrix, V1 is the gradient vector of the concentrated liquid, V2 is the texture feature vector of the concentrated liquid, V2 T is the transposed vector of the concentrated liquid texture feature vector, the concentrated liquid gradient vector and the concentrated liquid texture feature vector are both in column vector form, w1, w2, w3 and α are weight hyperparameters, represents vector multiplication, ⊙ represents point multiplication by position, represents the difference by position, Represents the implicit similarity factor loss function value of the common manifold.

10. A system for preparing a granular preparation of Evodia rutaecarpa, characterized in that: include: The first decoction module is used to add water and decoct the Evodia rutaecarpa slices for the first time to obtain a decoction liquid and a medicinal residue; A second decoction module is used to add water to the medicinal residue and decoct it for the second time to obtain a second decoction liquid; A decompression and concentration module, used for mixing the first decoction liquid and the second decoction liquid and concentrating them under reduced pressure to obtain a concentrated liquid, wherein the relative density of the concentrated liquid is within a predetermined range; a spray drying module, used for spray drying the concentrated solution to obtain Evodia rutaecarpa spray-dried powder; and The granulation module is used for granulating the spray-dried powder of Evodia rutaecarpa to obtain the Evodia rutaecarpa formula granules.

Images