CN116000895B - Quality detection robot and method for traditional Chinese medicine pharmacy process based on deep learning - Google Patents
Quality detection robot and method for traditional Chinese medicine pharmacy process based on deep learning Download PDFInfo
- Publication number
- CN116000895B CN116000895B CN202310311070.8A CN202310311070A CN116000895B CN 116000895 B CN116000895 B CN 116000895B CN 202310311070 A CN202310311070 A CN 202310311070A CN 116000895 B CN116000895 B CN 116000895B
- Authority
- CN
- China
- Prior art keywords
- detection
- sample
- chinese medicine
- traditional chinese
- box body
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Abstract
The invention discloses a quality detection robot and a quality detection method for a traditional Chinese medicine pharmaceutical process based on deep learning, and relates to the technical field of traditional Chinese medicine quality detection; the detection mechanism is fixed on the running mechanism and comprises a box body, a detection pool is arranged on the box body and is connected with the cavity through a pipeline, a relative densimeter interface is arranged in the cavity, and the relative densimeter interface is connected with the sample inlet through a pipeline. The detection method comprises the steps of presetting a working path of the robot; starting an operating mechanism and executing corresponding action commands; starting a detection flow; measuring the parameters of the medicine by adopting a high performance liquid chromatography; and (5) establishing a quantitative correction model. The material quality detection structure and the material quality detection process in the traditional Chinese medicine pharmacy process reduce the labor intensity and the personnel cost in the pharmacy process, and avoid the potential safety hazard and the quality risk brought by personnel operation to a certain extent.
Description
Technical Field
The invention belongs to the technical field of traditional Chinese medicine quality detection, and particularly relates to a traditional Chinese medicine pharmaceutical process quality detection robot and method based on deep learning.
Background
The quality detection of the traditional Chinese medicine production and manufacturing process is mainly aimed at feeding medicinal materials, intermediates and finished products, and detection of the medicinal materials in the pharmaceutical process is neglected, and one of the reasons is lack of process detection equipment suitable for production lines. At present, the mode of material detection in the pharmaceutical process of a traditional Chinese medicine extraction workshop mainly comprises two modes of on-line detection and off-line detection, wherein an installation interface of a detection instrument is reserved on pharmaceutical process equipment in advance, the method is not suitable for an established production line, and the problems of difficult instrument calibration, disassembly, cleaning and the like exist; the latter then needs the operating personnel to use different instruments to detect respectively after regularly taking a sample in production process, and complex operation also probably causes the potential safety hazard or introduces the quality risk. Therefore, there is a need to develop a set of pharmaceutical process detection devices that have autonomous mobile, autonomous positioning, autonomous sampling, autonomous detection capabilities and are adapted to the extraction shop environment. The robot technology is applied to the traditional Chinese medicine pharmacy field at present, such as unpacking mechanical arms, material transport AGV trolleys and the like, but a technical scheme for intelligent detection of an extraction workshop production line is not yet known.
The Chinese patent with publication number of CN113720730A discloses a device and a method for automatically detecting the yield of a traditional Chinese medicine extracting tank, wherein the device for automatically detecting the yield of the traditional Chinese medicine extracting tank comprises an extracting tank body, and the middle part of the extracting tank body is connected with a liquid extracting pipe; a reflux pipe is arranged at the bottom of the extraction tank body, and the surface of the reflux pipe is porous; the liquid extracting pipe extracts liquid from the extracting tank body into the reflux pipe; the bottom of the extraction tank body is provided with a sampling tube, and liquid of the sampling tube enters a density detection device for detection. However, the patent cannot perform a series of operations including autonomous movement, autonomous positioning, autonomous sampling and autonomous detection, thereby reducing the working efficiency.
Therefore, it is necessary to design a device and a method with simple structure, simple operation and low quality risk.
Disclosure of Invention
Aiming at the problems of complex operation, potential safety hazard and high quality risk existing in the traditional Chinese medicine preparation process material detection mode in the prior art, the invention provides a traditional Chinese medicine preparation process quality detection robot and a method based on deep learning.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows:
a quality detection robot for Chinese medicine preparation process based on deep learning comprises a detection mechanism, an execution mechanism and a travelling mechanism;
the actuating mechanism is fixed on the travelling mechanism and is connected with the detection mechanism in a communication mode;
the detection mechanism is fixed on the travelling mechanism and comprises a box body, a detection pool is arranged on the box body and is connected with the cavity through a pipeline, a densimeter interface is arranged in the cavity, and the densimeter interface is connected with the sample inlet through a pipeline.
Based on the above technical scheme, still further, running gear includes the AGV chassis, and installs the backup pad on the AGV chassis, installs the box in the backup pad, is equipped with actuating mechanism on the box, still is equipped with between backup pad and the AGV chassis and surveys the module.
Based on the technical scheme, still further, be equipped with the lift groove on the box, the outside extension of lift groove has lift platform, is equipped with detection mechanism on the lift platform.
Based on the above technical scheme, still further, the detection module is a laser radar.
Based on the technical scheme, further, the actuating mechanism is a six-axis mechanical arm, and the end part of the six-axis mechanical arm is provided with a clamping jaw and a positioning module.
Based on the technical scheme, the quality detection robot for the traditional Chinese medicine pharmacy process is further provided with instruments and meters which are arranged in the detection mechanism box body and include, but are not limited to, densitometers, PH meters, conductivity meters and spectrometers.
Based on the technical scheme, further, a first interface and a second interface are arranged on the detection pool, the conductivity meter or the PH meter is connected with the first interface, and the spectrometer is connected with the second interface.
Based on the above technical scheme, still further, still be equipped with lotion liquid storage pot and recovery mouth on the box, retrieve inside the mouth intercommunication box, lotion liquid storage pot is taken out lotion to the sample inlet through pipeline and pump for wash detection pond and pipeline after every sample detects.
A quality detection method of a traditional Chinese medicine pharmaceutical process based on deep learning comprises the following steps:
step S1: presetting a working path of a robot;
step S2: starting an operating mechanism and executing action commands of sampling, sample feeding and container recovery;
step S3: after the detection mechanism receives the sample poured into the sample inlet by the operation mechanism, a detection flow is started;
step S4: measuring the parameters of the medicine by adopting a high performance liquid chromatography;
step S5: and (5) establishing a quantitative correction model to finish detection.
Based on the above technical solution, in step S2, after the operating mechanism is powered on, a self-reset of the clamping jaw is performed, and the coded value in the register address starts to be continuously read.
Based on the technical scheme, further, after sampling, sample feeding, sample introduction and container recovery operation actions are completed according to the read information, a contracted code is written into a register address, an instruction is sent to a running mechanism and a detection mechanism through a modbus protocol, then the code of the register address is continuously read until a contracted code value sent by an operation mechanism is read, and then the next sampling, sample feeding, sample introduction and container recovery operation is executed.
Based on the above technical solution, the detecting process in step S3 further includes the following steps:
step S31: opening a pipeline between the sample inlet and the cavity, pumping the sample into a relative densimeter interface in the cavity, and measuring the relative density of the sample;
step S32: opening a pipeline between the cavity and the detection tank, pumping a sample into the detection tank from a relative densimeter interface, measuring the PH value or the conductivity of the sample, and collecting spectral data by a spectrometer;
step S33: the sample in the detection cell is discharged into the box body.
Based on the above technical scheme, further, the detection flow further includes:
step S34: sequentially pumping the washing liquid in the washing liquid storage tank into a sample inlet, a cavity and a detection pool to finish the cleaning of the detection mechanism;
step S35: and writing the appointed coding value into the register address to finish the detection flow.
Based on the above technical solution, further, the process of establishing the quantitative correction model in step S5 includes the following steps:
step S51: establishing a VAE model by using the unlabeled spectrum data;
step S52: constructing a spectrum matrix by using a plurality of continuous labeled spectrums taking sample collection time as a center;
step S53: and after the features of the sample spectrum are extracted by using the VAE model, an input matrix is generated, and the quality index of the sample is taken as an output training CNN model to predict the quality index content.
Compared with the prior art, the invention has the following beneficial effects:
(1) According to the invention, materials in the equipment are detected in time in the pharmaceutical process, the material change condition is known, the transparency degree of the pharmaceutical process of the traditional Chinese medicine is increased, the process state is timely judged, the quality problem in the production process is timely found, the quality inspection efficiency of the traditional Chinese medicine is improved, and the quality inspection cost is reduced;
(2) The invention integrates the instruments and meters for detecting the quality of various materials, can realize multi-dimensional quality evaluation and improves the detection efficiency;
(3) The invention combines two deep learning algorithms of VAE and CNN for the first time to construct a spectrum quantitative correction model, and can be used for establishing a rapid detection method with excellent prediction performance;
(4) The robot provided by the invention replaces operators to detect the quality of materials in the traditional Chinese medicine preparation process, reduces the labor intensity and the personnel cost in the preparation process, and avoids the potential safety hazard and the quality risk brought by personnel operation to a certain extent.
Drawings
FIG. 1 is a front view of a robot of the present invention without a detection mechanism;
FIG. 2 is a block diagram of the invention without the detection mechanism;
FIG. 3 is a schematic diagram of the structure of the detection mechanism of the present invention;
FIG. 4 is a flow chart of the detection method of the present invention;
FIG. 5 is a diagram of the working node of the running gear of embodiment 2 of the present invention;
FIG. 6 is a schematic diagram showing the ginkgolic acid content of the sample in the concentration process measured by the high performance liquid phase method in example 2 of the present invention;
FIG. 7 is a schematic structural diagram of a quantitative calibration model according to embodiment 2 of the present invention;
description of the drawings: 1. an operating mechanism; 11. a positioning module; 12. a clamping jaw; 2. a walking mechanism; 21. a detection module; 22. a case; 23. a support plate; 24. a lifting platform; 3. a detection mechanism; 31. a detection pool; 32. a sample inlet; 33. a washing liquid storage tank; 34. a relative densitometer interface; 35. a first interface; 36. a second interface; 37. recovering the mouth; 38. a box body.
Detailed Description
It is to be noted that the raw materials used in the present invention are all common commercial products, and the sources thereof are not particularly limited.
The following raw material sources are exemplary illustrations:
example 1
As shown in fig. 1, the quality detection robot for the traditional Chinese medicine pharmacy process based on deep learning comprises a detection mechanism 3, an execution mechanism and a travelling mechanism 2; the travelling mechanism 2 is used for drawing a workshop map and moving between working points according to a preset route in the map; the operating mechanism 1 is used for positioning an execution part of the pharmaceutical equipment and completing a specified operation; the detection mechanism 3 is used for realizing multidimensional rapid detection of the collected pharmaceutical process material samples. The operating mechanism 1 and the detecting mechanism 3 are fixed on the travelling mechanism 2 through fasteners and can communicate based on industrial communication protocols, including but not limited to Modbus, TCP/IP, RS-232, RS-485 and the like; the send instruction includes, but is not limited to, writing a specified code to the agreed register address, triggering a physical key, etc. After the travelling mechanism 2 moves to the target point on the map, an instruction is sent to the operating mechanism 1, and the operating mechanism 1 executes corresponding operation. After the operation of the operation mechanism 1 is completed, an instruction is sent to the detection mechanism 3, and the detection mechanism 3 starts to collect data of the detection instrument. After the detection of the sample is completed by the detection mechanism 3, an instruction is sent to the traveling mechanism 2, and the traveling mechanism 2 moves to the next target point or charging point along the set path.
Specifically, the travelling mechanism 2 comprises an AGV chassis, a supporting plate 23 is arranged on the AGV chassis, a box 22 is arranged on the supporting plate 23, an executing mechanism is arranged on the box 22, a detection module 21 is further arranged between the supporting plate 23 and the AGV chassis, and the detection module 21 can be a laser radar which can realize the functions of map drawing, autonomous obstacle avoidance, autonomous charging and the like; the executing mechanism is connected with the detecting mechanism 3 in a communication mode; the detection mechanism 3 is fixed on the travelling mechanism 2, as shown in fig. 2 and 3, specifically, a lifting groove is formed on the box 22, a lifting platform 24 is outwardly extended from the lifting groove, and the detection mechanism 3 is arranged on the lifting platform 24. And detection mechanism 3 includes box 38, this box 38 is used for waste liquid storage, the sampling container of retrieving of keeping in, be equipped with detection pond 31 on the box 38, detection pond 31 passes through the pipe connection cavity, the densimeter interface has been seted up in the cavity, be connected densimeter interface and sample inlet 32 through the pipeline, the inside container recovery jar that is equipped with of box 38, detection instrument and valve, and the container is retrieved jar, detection instrument and valve between mutually independent setting, actuating mechanism is six mechanical arms, clamping jaw 12 and positioning module 11 are installed to six mechanical arm's tip, this positioning module 11 can be servo positioning camera. Wherein, all be equipped with corresponding control valve and pump on each pipeline, through the switch of control corresponding valve and the start-stop of pump, can realize the control on the corresponding pipeline. Detection instrumentation includes, but is not limited to, densitometers, PH meters, conductivity meters, spectrometers. The detection cell 31 is provided with a first interface 35 and a second interface 36, the conductivity meter or the PH meter is connected with the first interface 35, and the spectrometer is connected with the second interface 36, wherein the same interface can not be connected with two parts at the same time, but can be replaced with different parts. The box 38 is also provided with a lotion liquid storage tank 33 and a container recovery port 37, the recovery port 37 is communicated with the container recovery tank in the box 38, and the lotion liquid storage tank 33 pumps the lotion to the sample inlet 32 through a pipeline for cleaning the detection tank 31 and the pipeline after each sample detection.
Example 2
The quality detection method of the traditional Chinese medicine pharmaceutical process based on deep learning as shown in fig. 4 comprises the following steps:
step S1: presetting a working path of a robot; specifically, a laser radar is installed on an AGV chassis, a charging port and a button for controlling sudden stop are arranged on the AGV chassis, after a path is preset, a debugging person operates a robot travelling mechanism 2 to move in a workshop in a manual mode, and the robot senses the workshop environment through the arranged laser radar and autonomously draws a workshop map; as shown in fig. 5, after the mapping is completed and started, the robot enters a designated point in the map and records, wherein A, B, C, D, E, F, G, H, I represents a point, a point is set as a robot charging point, a point is set as a robot stopping point, points F, I and D are respectively extraction, concentration and chromatography sampling working points, and the rest points are path planning points. After the point position recording is completed, the robot walking path planning is realized by connecting the point positions and setting the direction. After the path planning is completed, a task flow is set, firstly, a working point position is selected, then, a contracted code is written into a register address 9000, and an instruction is sent to an operating mechanism 1 through a modbus protocol, wherein the code written into an F number point position is 1, the code written into an I number point position is 2, and the code written into a D number point position is 3, and the codes 1, 2 and 3 are contracted codes in the embodiment, namely, contracted codes. The reading of the code of register address 9000 is then continued until the contract code sent by operating mechanism 1 is read and returned to the stop point along the planned path.
Step S2: starting the operating mechanism 1; and executing the action commands of sampling, sample feeding, sample introduction and container recovery;
step S3: after the detection mechanism 3 receives the sample poured into the sample inlet 32 by the operation mechanism 1, a detection flow is started;
specifically, the steps S2 and S3 are that, after the operating mechanism 1 is powered on, according to a self-checking program set in advance for the clamping jaw 12, the self-resetting of the clamping jaw 12 is performed once, and the continuous reading of the code value in the register address 9000 is started, and when the contract code is read, the operation is started. In the figure, an SETx node is a register read-write node, a Grip_Shift_V21 node is a clamping jaw 12 operation node, tq, ns and cx are respectively extraction, concentration and chromatography sampling valve servo positioning nodes, a Pxxx node is a mechanical arm positioning node, and the rest nodes are nodes of logic judgment, overtime, error and the like. After the operations such as sampling, sample feeding, and container recycling are completed, the agreed code is written into the register address 9000, and an instruction is sent to the travelling mechanism 2 and the detection mechanism 3 through the modbus protocol, and in this embodiment, the agreed code value is set to be 4. The reading of the code of register address 9000 is then continued until the next operation is performed after reading the agreed code value sent by operating mechanism 1.
The detection mechanism 3 starts the detection flow after receiving the sample poured into the sample inlet 32 by the operation mechanism 1 and reading the agreed code value of the register address 9000. The specific detection flow is as follows:
step S31: a control valve between the sample inlet 32 and the cavity is opened, and the sample is pumped into a relative densitometer interface 34 in the cavity to measure the relative density of the sample;
step S32: a control valve on a pipeline between the cavity and the detection tank 31 is opened, the relative densimeter interface 34 is communicated with the detection tank 31, a sample is pumped into the detection tank 31 from the relative densimeter interface 34 to measure the pH value or the conductivity, at the moment, the spectrometer is in a continuous collection state, and after the sample enters the detection tank 31, the spectrometer automatically collects the spectral data of the sample;
step S33: the waste port valve below the detection cell 31 is opened and the sample is discharged into the tank 38.
After the above process is completed, the following steps are further performed:
step S34: sequentially pumping the washing liquid in the washing liquid storage tank 33 into the sample inlet 32, the cavity and the detection pool 31 to finish the cleaning of the detection mechanism 3; the reason why the detection mechanism 3 is cleaned is that the sample remains after passing through the pipe and the detection cell 31, and if the cleaning is not performed, the sample remaining last time is mixed into the sample to be detected next time, and the detection result is inaccurate.
Step S35: and writing the appointed coding value into the register address 9000 to finish the detection flow.
Step S4: measuring the parameters of the medicine by adopting a high performance liquid chromatography;
step S5: and (5) establishing a quantitative correction model to finish detection.
Specifically, the establishment process of the quantitative correction model comprises the following steps:
step S51: establishing a VAE model by using the unlabeled spectrum data; the label-free spectrum data are collected by a spectrometer, and only the corresponding reference value of the sample corresponding to the spectrum is not measured, namely the sample is not further analyzed by high performance liquid chromatography and other methods;
step S52: constructing a spectrum matrix by using a plurality of continuous labeled spectrums taking sample collection time as a center; the robot starts from the initial moment of the pharmaceutical process, samples at fixed intervals and detects the collected samples, and the sample collection moment is the time node of the robot for collecting the samples; the number of the several tagged spectrums can be preferably 8-32; the method is the same as the method for acquiring the non-tag spectrum, and the sample corresponding to the part of the spectrum with the tag spectrum is measured by a reference value, namely, the content of the index component is measured by a high performance liquid chromatography method and the like; and continuous multiple spectrums are adopted as model input by taking sample collection time as a center, so that the influence of abnormal spectrums on a model prediction result can be reduced.
Step S53: and after the features of the sample spectrum are extracted by using the VAE model, an input matrix of 8-32 multiplied by 8-32 is generated, and the sample quality index is taken as output to train the CNN model, so that the quality index content is predicted. The output of the VAE model is the input of the CNN model; predicting by using a CNN model, wherein the function of the model is to predict the content of the index component through the spectrum characteristic; the whole model is used for predicting the content of the index components, and the content model is predicted to be built. The different quality indexes are obtained by detecting through different reference methods, such as detecting the content of index components by HPLC, and obtaining the relative density by measuring by a specific gravity method.
Example 3
Based on the embodiment 1 and the embodiment 2, the quality indexes of different samples are different, and taking ginkgolic acid as an example, the quality index in the embodiment is ginkgolic acid content, and ginkgolic acid has proved to have potential sensitization, mutation effect and strong cytotoxicity, can cause serious allergic reaction, gene mutation and nerve injury, and cause adverse reactions such as nausea, heartburn, anaphylactic shock, spasm, nerve paralysis and the like, and needs to be closely monitored in the production process. The embodiment realizes the real-time monitoring of ginkgolic acid in the concentration process, and on the basis of the embodiment 2, the detection process is as follows: the robot collects samples of the ginkgo leaf concentration process autonomously: in the concentration process, about 20 parts of samples mL are collected every 5 minutes, and 75 parts of concentrated liquid medicine samples are collected. The robot pours the sample into the detection module, starts the detection flow, collects the Raman spectrum data, and sets the integration time to be 1s. The reference value of ginkgolic acid is determined by adopting a high performance liquid chromatography, and the chromatographic conditions are as follows: the column was Agilent Zorbax SB-C18 (4.6X1250 mm,5 μm). Mobile phase is 3% aqueous acetic acid (phase a): pure methanol (phase B) =8:92; the flow rate was 1.0. 1.0 mL/min. The column temperature was 30℃and the sample injection amount was 10. Mu.L, the ultraviolet detection wavelength was 310 and nm, and the measurement results of the reference values are shown in FIG. 6.
And then a quantitative correction model is established by adopting an algorithm combining a variational self-encoder VAE and a convolutional neural network CNN and is used for rapidly determining the ginkgolic acid content of a sample, and the model structure is shown in figure 7. Setting the number of nodes in the coding network to 256, and setting the length of hidden variable (coding) { c1, c2, … …, c16} to 16; the input Raman spectrum outputs mean vectors { m1, m2, … …, m16} and variance vectors { sigma 1, sigma 2, … …, sigma 16} which are equal to the length of the hidden variables after passing through the coding network, and simultaneously generates a vector { e1, e2, … …, e16} which accords with standard normal distribution as Gaussian noise; the CNN model contains 2 convolutional layers, 2 pooled layers, and 2 fully connected layers, the parameters of each layer are shown in table 1. In the model training process, mean square error is used as loss, and a strategy of dropout and early stopping is adopted to prevent overfitting.
TABLE 1
The trained model is evaluated by three indexes of a test set, namely a decision coefficient (R2), a Root Mean Square Error (RMSE) and a Residual Prediction Deviation (RPD), and compared with the traditional PLSR, and the evaluation and comparison results are shown in a table 2, so that the prediction performance of the ginkgolic acid content quantitative correction model built by the embodiment is obviously superior to that of the traditional PLSR model, and the process monitoring requirement is met.
TABLE 2
Finally, it should be noted that the above description is only for illustrating the technical solution of the present invention, and not for limiting the scope of the present invention, and that the simple modification and equivalent substitution of the technical solution of the present invention can be made by those skilled in the art without departing from the spirit and scope of the technical solution of the present invention.
Claims (2)
1. A quality detection method of traditional Chinese medicine pharmacy process based on deep learning, which uses a quality detection robot of traditional Chinese medicine pharmacy process based on deep learning, is characterized in that,
the detection robot comprises a detection mechanism, an execution mechanism and a travelling mechanism;
the actuating mechanism is fixed on the travelling mechanism and is connected with the detection mechanism in a communication mode;
the detection mechanism is fixed on the travelling mechanism and comprises a box body, a detection pool is arranged on the box body and is connected with the cavity through a pipeline, a relative densimeter interface is arranged in the cavity, and the relative densimeter interface is connected with the sample inlet through a pipeline;
the traveling mechanism comprises an AGV chassis, a supporting plate is arranged on the AGV chassis, and a detection module is arranged between the supporting plate and the AGV chassis; the box body is arranged on the supporting plate, the executing mechanism is a six-axis mechanical arm, and the clamping jaw and the positioning module are arranged at the end part of the six-axis mechanical arm;
the box body is provided with a lifting groove, the lifting groove extends outwards to form a lifting platform, and the lifting platform is provided with a detection mechanism;
the quality detection robot for the traditional Chinese medicine pharmacy process is also provided with an instrument and meter which is arranged in the detection mechanism box body and comprises a densimeter, a PH meter, a conductivity meter and a spectrometer; the detection pool is provided with a first interface and a second interface, the conductivity meter or the PH meter is connected with the first interface, and the spectrometer is connected with the second interface;
the box body is also provided with a lotion liquid storage tank and a recovery port, the recovery port is communicated with the inside of the box body, and the lotion liquid storage tank pumps lotion to the sample inlet through a pipeline;
the method comprises the following steps:
step S1: presetting a working path of a robot;
step S2: after the operating mechanism is electrified, the self-resetting of the clamping jaw is firstly executed once, the code in the register address is continuously read, the operating mechanism is started, and the action commands of sampling, sample feeding, sample introduction and container recovery are executed;
step S3: after the detection mechanism receives the sample poured into the sample inlet by the operation mechanism, a detection flow is started;
the detection process in step S3 includes the following steps:
step S31: opening a pipeline between the sample inlet and the cavity, and measuring the relative density of the sample;
step S32: opening a pipeline between the cavity and the detection pool, measuring the PH value or the conductivity of the sample, and collecting spectral data by a spectrometer;
step S33: discharging the sample in the detection cell into the box body;
the detection flow also comprises:
step S34: sequentially pumping the washing liquid in the washing liquid storage tank into a sample inlet, a cavity and a detection pool to finish the cleaning of the detection mechanism;
step S35: writing a contract coding value into a register address to finish a detection flow;
step S4: measuring the parameters of the medicine by adopting a high performance liquid chromatography;
step S5: establishing a quantitative correction model to finish detection;
the establishment process of the quantitative correction model comprises the following steps:
step S51: establishing a VAE model by using the unlabeled spectrum data;
step S52: constructing a spectrum matrix by using a plurality of continuous labeled spectrums taking sample collection time as a center;
step S53: and after the features of the sample spectrum are extracted by using the VAE model, an input matrix is generated, and the quality index of the sample is taken as an output training CNN model to predict the quality index content.
2. The method for detecting the quality of the traditional Chinese medicine pharmaceutical process based on deep learning according to claim 1, wherein after sampling, sample feeding, sample introduction and container recovery operations are completed according to the read information, a contracted code is written into a register address, an instruction is sent to a running mechanism and a detection mechanism through a modbus protocol, then the code of the register address is continuously read until a contracted code value sent by the operation mechanism is read, and then the next sampling, sample feeding, sample introduction and container recovery operations are executed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310311070.8A CN116000895B (en) | 2023-03-28 | 2023-03-28 | Quality detection robot and method for traditional Chinese medicine pharmacy process based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310311070.8A CN116000895B (en) | 2023-03-28 | 2023-03-28 | Quality detection robot and method for traditional Chinese medicine pharmacy process based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116000895A CN116000895A (en) | 2023-04-25 |
| CN116000895B true CN116000895B (en) | 2023-06-23 |
Family
ID=86019648
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310311070.8A Active CN116000895B (en) | 2023-03-28 | 2023-03-28 | Quality detection robot and method for traditional Chinese medicine pharmacy process based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116000895B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106826780A (en) * | 2017-02-20 | 2017-06-13 | 成都科创谷科技有限公司 | A kind of Chinese medicine pharmacy's special manipulator |
| CN107377413A (en) * | 2017-08-18 | 2017-11-24 | 南京工业职业技术学院 | Infrared fruit internal quality grading robot |
| CN109883548A (en) * | 2019-03-05 | 2019-06-14 | 北京理工大学 | The Encoding Optimization of the spectrum imaging system of neural network based on optimization inspiration |
| CN112926830A (en) * | 2021-01-25 | 2021-06-08 | 南通市第一人民医院 | Real-time-monitorable artificial intelligent quality monitoring system for traditional Chinese medicine decoction storage and control method |
| CN113030008A (en) * | 2021-02-25 | 2021-06-25 | 南京中医药大学 | Near-infrared online quality detection method for cattail pollen charcoal processed product |
| CN113702328A (en) * | 2021-08-20 | 2021-11-26 | 广东省惠州市石油产品质量监督检验中心 | Method, device, equipment and storage medium for analyzing properties of product oil |
Family Cites Families (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8428889B2 (en) * | 2010-10-07 | 2013-04-23 | Thermo Finnigan Llc | Methods of automated spectral peak detection and quantification having learning mode |
| US10664716B2 (en) * | 2017-07-19 | 2020-05-26 | Vispek Inc. | Portable substance analysis based on computer vision, spectroscopy, and artificial intelligence |
| CN108875913B (en) * | 2018-05-30 | 2021-09-17 | 江苏大学 | Tricholoma matsutake rapid nondestructive testing system and method based on convolutional neural network |
| CN109446970B (en) * | 2018-10-24 | 2021-04-27 | 西南交通大学 | Transformer substation inspection robot road scene recognition method based on deep learning |
| CN109329501A (en) * | 2018-10-29 | 2019-02-15 | 江苏大学 | A kind of intelligent preparation method of instant fermented tea powder |
| US20220057804A1 (en) * | 2018-12-27 | 2022-02-24 | Honda Motor Co., Ltd. | Path determination method |
| CN111974704A (en) * | 2020-08-14 | 2020-11-24 | 东北大学秦皇岛分校 | Garbage classification detection system and method based on computer vision |
| CN111966102B (en) * | 2020-08-18 | 2023-04-14 | 昆山三一环保科技有限公司 | Intelligent robot sampling system for automatic detection and control method thereof |
| US20220121940A1 (en) * | 2020-10-20 | 2022-04-21 | The Regents Of The University Of California | Device and method for neural-network based on-chip spectroscopy using a plasmonic encoder |
| CN113588590B (en) * | 2021-08-11 | 2024-04-16 | 苏州泽达兴邦医药科技有限公司 | Traditional Chinese medicine extraction process quality control method based on data mining |
| CN113702531A (en) * | 2021-08-28 | 2021-11-26 | 江西中医药大学 | Online monitoring method and device for traditional Chinese medicine extraction process endpoint |
-
2023
- 2023-03-28 CN CN202310311070.8A patent/CN116000895B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106826780A (en) * | 2017-02-20 | 2017-06-13 | 成都科创谷科技有限公司 | A kind of Chinese medicine pharmacy's special manipulator |
| CN107377413A (en) * | 2017-08-18 | 2017-11-24 | 南京工业职业技术学院 | Infrared fruit internal quality grading robot |
| CN109883548A (en) * | 2019-03-05 | 2019-06-14 | 北京理工大学 | The Encoding Optimization of the spectrum imaging system of neural network based on optimization inspiration |
| CN112926830A (en) * | 2021-01-25 | 2021-06-08 | 南通市第一人民医院 | Real-time-monitorable artificial intelligent quality monitoring system for traditional Chinese medicine decoction storage and control method |
| CN113030008A (en) * | 2021-02-25 | 2021-06-25 | 南京中医药大学 | Near-infrared online quality detection method for cattail pollen charcoal processed product |
| CN113702328A (en) * | 2021-08-20 | 2021-11-26 | 广东省惠州市石油产品质量监督检验中心 | Method, device, equipment and storage medium for analyzing properties of product oil |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116000895A (en) | 2023-04-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109407654B (en) | Industrial data nonlinear causal analysis method based on sparse deep neural network | |
| CN109709057A (en) | Water quality indicator prediction model construction method and water quality indicator monitoring method | |
| CN110996280B (en) | RFID indoor positioning fingerprint database updating system and method | |
| Yuan et al. | A spatial-temporal LWPLS for adaptive soft sensor modeling and its application for an industrial hydrocracking process | |
| CN110322014A (en) | A kind of finished cement specific surface area prediction technique based on BP neural network | |
| CN116000895B (en) | Quality detection robot and method for traditional Chinese medicine pharmacy process based on deep learning | |
| CN102968664B (en) | A kind of mobile robot's variable farm chemical applying Using Intelligent Decision-making Method and system | |
| CN108647485B (en) | Method, system, medium and equipment for measuring carbon deposition of catalyst in fluid catalytic cracking | |
| CN110597065A (en) | Adaptive control system for powder feeding type laser additive | |
| Cui et al. | Deep learning methods for atmospheric PM2. 5 prediction: A comparative study of transformer and CNN-LSTM-attention | |
| AU2020102518A4 (en) | A method of gas sensor calibration based on linear optimization | |
| CN102289718A (en) | Method for softly measuring height of mud layer in settlement process of red mud | |
| CN116738866A (en) | Instant learning soft measurement modeling method based on time sequence feature extraction | |
| Sun et al. | Optimization of extraction process based on neural network | |
| Yang et al. | Application of fuzzy neural network PID algorithm in oil pump control | |
| CN110763830A (en) | Method for predicting content of free calcium oxide in cement clinker | |
| CN116659568A (en) | On-spot environmental monitoring system of boats and ships application workshop | |
| WO2022241883A1 (en) | Pollution source recognition method and apparatus based on three-dimensional liquid chromatography fingerprint | |
| CN114414648A (en) | Automatic potentiometric titration method and system based on machine learning | |
| Klemencic et al. | Application of algorithms and machine learning methods in pharmaceutical manufacture | |
| Kurz et al. | Observation and control of the beer fermentation using cognitive methods | |
| CN114446410A (en) | XGboost-based inorganic arsenic content prediction method, device and medium | |
| CN208606794U (en) | A kind of oil-immersed power transformer fault detection system | |
| CN111443159A (en) | Gas-sensitive-gas chromatography multi-source sensing and electronic nose instrument on-line detection method | |
| CN111812041A (en) | Portable water body COD (chemical oxygen demand) measuring system and method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |