Semiconductor liquid medicine accurate temperature control system based on deep learning

Abstract

The invention relates to a semiconductor liquid medicine accurate temperature control system based on deep learning, which comprises a heat exchanger module connected with a chemical liquid medicine and a neural network learning device, and is characterized in that the heat exchanger module is connected with a control module, the control module is connected with an artificial intelligence module, and the artificial intelligence module is connected with the neural network learning device. The neural network learning device is connected with the artificial intelligence module. A heat exchanger module for heating or cooling a liquid medicine supplied from a liquid medicine tank, a control module for performing cooling or heating of the heat exchanger module, and a temperature prediction module based on deep learning for predicting the discharge temperature of the liquid medicine discharged from the heat exchanger module are provided, so that the temperature in a liquid medicine tube can be monitored and controlled in real time. The temperature of the liquid medicine is accurately controlled while the transmission efficiency of the liquid medicine is improved, and the smooth reaction process of the liquid medicine is kept.

Description

Technical Field

[0001] The present invention belongs to the technical field of semiconductor drug supply systems, and specifically relates to a semiconductor drug solution precise temperature control system based on deep learning. Background Art

[0002] Nano-scale semiconductors are essential practical equipment for display processes. The central chemical supply system is a central chemical supply system that performs the supply and mixing functions of fine chemicals. Nano-scale semiconductor precision processes are the core of manufacturing facilities. It is one of the essential equipment that is installed first along with the clean room.

[0003] Generally, in semiconductor processes and liquid crystal display (LCD) processes, substrate processing devices are used to perform surface treatment on substrates using various chemicals. For example, the substrate processing device performs etching treatment, cleaning treatment, etc. on the substrate, and in these substrate treatments, acidic solutions such as potassium hydroxide, sodium hydroxide, sodium hydroxide, or alkaline solutions such as potassium hydroxide, sodium hydroxide, or any one of them or a mixture thereof, as well as various types of chemical solutions are used.

[0004] By using these different types of liquids to cause chemical reactions, unnecessary substances on the substrate are removed or cleaned. In the process of processing the substrate, the temperature of the liquid sprayed onto the substrate is a major factor in the process, that is, during the etching and cleaning process, keeping the temperature of the liquid constant is essential for the smooth progress of the chemical reaction. Summary of the invention

[0005] The purpose of the present invention is to solve the above-mentioned problems. The present application proposes a semiconductor liquid precision temperature control system based on deep learning, which is provided with a heat exchanger module for heating or cooling the liquid supplied from the liquid tank, a control module for executing cooling or heating of the heat exchanger module, and a temperature prediction module based on deep learning for predicting the discharge temperature of the liquid discharged from the heat exchanger module. The system can monitor and control the temperature inside the liquid tube in real time, improve the liquid transfer efficiency, and accurately control the liquid temperature, thereby maintaining the smooth progress of the liquid reaction process.

[0006] A semiconductor liquid precise temperature control system based on deep learning, comprising a heat exchanger module connected to the chemical liquid, a neural network learning device, the heat exchanger and the control module are interconnected, the control module and the artificial intelligence module are interconnected, and the neural network learning device is connected to the artificial intelligence module: The heat exchanger module comprises a plurality of unit heat exchangers, a first heat sink, a second heat sink, and a water-cooling block, each of the unit heat exchangers comprises a pipe temperature measuring sensor for measuring the temperature of a medicine liquid pipe at a specified position, a first heat sink and a second heat sink are respectively arranged on both sides of the heat exchanger, the first heat sink and the second heat sink are formed as a square plate having a pipe with a diameter greater than an outer diameter of the medicine liquid pipe, the inlet end of the medicine liquid pipe is arranged in the first heat sink, and the outlet end is arranged in the second heat sink, a water-cooling block is further arranged between the first heat sink and the second heat sink, a cooling water channel is arranged inside the water-cooling block, a thermoelectric element is arranged between the first heat sink and the water-cooling block, and between the second heat sink and the water-cooling block, the thermoelectric element is composed of a plurality of thermocouples, the thermocouple heats or cools the medicine liquid pipe at a specified position, and a thermoelectric element controller controls the direction of current flowing through the thermocouple; A plurality of T-shaped polytetrafluoroethylene connectors are provided on the medicine liquid tube between the first heat sink block and the second heat sink block, a plurality of pipeline temperature measurement sensors are provided around the T-shaped polytetrafluoroethylene connector, and the T-shaped polytetrafluoroethylene connector is connected to an internal temperature probe for measuring the temperature of the medicine in the medicine liquid tube.

[0007] Furthermore: the heat exchanger cools or heats the temperature of the medicine in the medicine pipe, and then sends the temperature measured by the pipeline temperature measurement sensor to the control module through the temperature value sending device; the artificial intelligence module learns according to the training data to determine the temperature control model, and the control module predicts the temperature of the medicine to be supplied to the medicine tank according to the temperature control model provided by the artificial intelligence module, and performs cooling or heating control according to the predicted temperature.

[0008] Furthermore: the first heat dissipation block and the second heat dissipation block are respectively provided with six straight-through parts A inside along the length direction, and multiple folding parts B outside, and the straight-through parts A are connected to each other in the vertical and horizontal directions through the folding parts B.

[0009] Furthermore: the shape of the water cooling block is a square plate corresponding to the first heat dissipation block and the second heat dissipation block.

[0010] Further: the neural network learning device is a server of the control system, which is completed based on the deep learning algorithm. The neural network learning device can send the learned facial generation model to the facial generation unit periodically or upon request, or update the installed facial generation model. The neural network learning device includes a communication unit, an input unit, a memory, a learning processor and a processor.

[0011] Preferably: the communication unit includes a wireless communication unit and an interface unit, and the communication unit is connected to the drone device through wired or wireless communication and sends or receives data thereto; The input unit acquires training data for model learning and uses the training model to obtain output input data; The memory stores a model learned by a learning processor or a neural network learning device, and the memory includes a model memory or a database, wherein the model memory stores a neural network model that is being learned or has been learned by the learning processor; the database is used to store input data obtained by an input unit, learning data used for model learning, and a learning history of the model; The learning processor uses training data or training sets to train the artificial neural network. The learning processor learns the artificial neural network by directly acquiring preprocessed data of input data acquired by the processor through an input unit, or by acquiring preprocessed input data stored in a database.

[0012] Preferably, the deep learning algorithm uses a recurrent neural network (RNN) and a long short-term memory (LTSM).

[0013] Preferably, the liquid medicine tube is made of fluororesin materials with strong chemical resistance and corrosion resistance, including PEA and polytetrafluoroethylene.

[0014] Preferably, the pipeline temperature measurement sensor is an RTD temperature sensor.

[0015] Preferably, the liquid medicine tube is made of fluororesin materials with strong chemical resistance and corrosion resistance, including PEA and polytetrafluoroethylene.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The present invention is provided with a heat exchanger module for heating or cooling the medicine supplied from the medicine tank; a control module for performing cooling or heating of the heat exchanger module; a temperature prediction module based on deep learning for predicting the discharge temperature of the medicine discharged from the heat exchanger module; in order to increase the heat transfer area with the heat sink, the medicine pipe is arranged in six straight-through forms inside the first and second heat sinks along the length direction of the heat sink, and a water cooling block is provided between the heat sinks, and a cooling water channel is provided inside the water cooling block. Moreover, since the thermoelectric element is installed between the heat sink and the water cooling block, the heat generated by the thermoelectric element can be cooled to prevent damage caused by thermal shock. Moreover, the medicine pipe is made of fluororesin material with strong chemical resistance and corrosion resistance to avoid reaction with the medicine. The present invention can monitor and control the temperature in the medicine pipe in real time, improve the efficiency of medicine transfer, and accurately control the temperature of the medicine, so as to keep the reaction process of the medicine going smoothly. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only for more clearly illustrating the embodiments of the present invention or the technical solutions in the prior art. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 It is a structural diagram of a liquid medicine temperature control device in the prior art; Figure 2 It is the overall framework diagram of the temperature control system of the present invention; Figure 3 A side view of a schematic structural diagram of a heat exchanger module of a temperature control system of the present invention; Figure 4 It is a rear view of the schematic structural diagram of the heat exchanger module of the temperature control system of the present invention; Figure 5 This is a state diagram after the heat sink block in the heat exchanger module of the temperature control system of the present invention is removed; Figure 6 A side view of a water cooling block and a thermoelectric element in a heat exchanger module of a temperature control system of the present invention; Figure 7 It is a schematic diagram of the structure of a heat exchanger module in a temperature control system of the present invention, in which a pipe temperature measurement sensor is removed from the heat exchanger module; Figure 8 This is the internal configuration diagram of the unit heat exchanger of the present invention; Fig. 9 Learning device connectivity graphs for neural networks; Fig.10 RNN and LATM model diagrams for AI module examples; In the figure: 1-heat exchanger module, 101-unit heat exchanger, 102-first heat sink, 103-second heat sink, 104-water cooling block, 105-thermoelectric element, 1051-thermoelectric element controller, 2-control module, 3-artificial intelligence module, 5-neural network learning device, 51-communication unit, 52-input unit, 53-memory, 531-model memory, 532-database, 54-learning processor, 55-processor, 6-inlet end, 7-outlet end, 8-T-shaped polytetrafluoroethylene connector, 9-internal temperature probe, 10-pipeline temperature measurement sensor, 11-temperature value sending device; A-straight part, B-folded part. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand and implement the technical solution of the present invention, the present invention is further described below in conjunction with specific embodiments. However, the embodiments are only intended to illustrate the present invention and not to limit the present invention.

[0020] like Figure 2 The semiconductor liquid precision temperature control system based on deep learning shown in the figure includes a heat exchanger module 1, a control module 2, an artificial intelligence module 3, and a neural network learning device 5 connected to the chemical liquid pipe. The heat exchanger module 1 and the control module 2 are interconnected, the control module 2 and the artificial intelligence module 3 are interconnected, and the neural network learning device 5 is connected to the artificial intelligence module 3.

[0021] The heat exchanger module 1 is used to heat or cool the liquid medicine in the liquid medicine pipe, the artificial intelligence module 3 learns according to the training data to determine the temperature control model, the control module 2 predicts the temperature of the liquid medicine to be supplied to the liquid medicine tank according to the temperature control model provided by the artificial intelligence module 3, and performs cooling or heating control according to the predicted temperature. The control module is an execution module that controls the working state of the heat exchange module, that is, cooling or heating.

[0022] like Figures 3 to 7 As shown, the heat exchanger module 1 includes a plurality of unit heat exchangers 101, a first heat sink 102, a second heat sink 103, and a water cooling block 104. Each of the unit heat exchangers 101 includes a pipe temperature measuring sensor 10 for measuring the temperature of a liquid medicine pipe at a specified position. The first heat sink 102 and the second heat sink 103 are respectively provided on both sides of the heat exchanger module 1. The first heat sink 102 and the second heat sink 103 are formed as a square plate having a pipe with a diameter greater than the outer diameter of the liquid medicine pipe. The inlet end 6 of the liquid medicine pipe is provided in the first heat sink 102, and the outlet end 7 is provided in the second heat sink 103. A water cooling block 104 is further provided between the first heat dissipation block 102 and the second heat dissipation block 103. The shape of the water cooling block 104 is a square plate corresponding to the first heat dissipation block 102 and the second heat dissipation block 103. A cooling water channel is provided inside the water cooling block 104 through which cooling water can flow. A thermoelectric element 105 is provided between the first heat dissipation block 102 and the water cooling block 104 and between the second heat dissipation block 103 and the water cooling block 104. The thermoelectric element 105 is composed of a plurality of thermocouples. The thermocouple heats or cools the medicine liquid tube at a specified position and controls the direction of the current flowing through the thermocouple. A thermoelectric element controller 1051, A plurality of T-shaped polytetrafluoroethylene connectors 8 are provided on the medicine liquid tube between the first heat sink block 102 and the second heat sink block 103, a plurality of pipeline temperature measurement sensors 10 are provided around the T-shaped polytetrafluoroethylene connector 8, and the T-shaped polytetrafluoroethylene connector 8 is connected to an internal temperature probe 9 for measuring the temperature of the medicine in the medicine liquid tube.

[0023] Preferably: in order to increase the heat transfer area between the first heat dissipation block and the second heat dissipation block, six straight-through parts A are respectively provided inside the first heat dissipation block 102 and the second heat dissipation block 103 along the length direction, and a plurality of folding parts B are provided outside, and the straight-through parts A are connected to each other in the vertical and horizontal directions through the folding parts B.

[0024] Preferably, the thermoelectric elements 105 in the unit heat exchanger 101 are preferably designed as a sealed structure capable of being individually insulated, and the thermoelectric elements 105 in each unit heat exchanger 101 do not affect each other.

[0025] Preferably: more than 20 thermoelectric elements are used in the control system, which are controlled individually and installed at different points of the liquid circulation flow of the medicine to collect a large amount of temperature data, and the data is processed by deep learning method.

[0026] In the present invention, a neural network learning device is used to train data. The neural network learning device 5 can communicate with at least one external device or terminal, and can replace the external device or help the external device to analyze or learn data to obtain results. It can periodically or upon request send the learned facial generation model to the facial generation unit 230, or update the installed facial generation model.

[0027] like Fig. 9 As shown, the neural network learning device 5 includes a communication unit 51, an input unit 52, a memory 53, a learning processor 54 and a processor 55.

[0028] The communication unit 51 includes a wireless communication unit and an interface unit. The communication unit 51 can be connected to a drone device or other devices through wired or wireless communication and send or receive data thereto.

[0029] The input unit 52 acquires training data for model learning and uses the training model to obtain output input data; the input unit 52 may also acquire unprocessed input data, at which point the learning processor 54 or the processor 55 may preprocess the acquired data to generate training data or preprocessed input data that may be input into model learning; The memory 53 stores the model learned by the learning processor or the neural network learning device, and the memory can divide the learned model into multiple versions and store them according to the learning time and learning progress, etc. The memory can store input data obtained from the input unit 52, learning data or training data for model learning, and the learning history of the model; it can also store unprocessed or processed input data.

[0030] The memory includes a model memory 531 or a database 532. The model memory 531 stores the neural network model that is being learned or has been learned by the learning processor, and when the model is updated through learning, the updated model is stored. The learned model can also be divided into multiple versions for storage according to learning time or learning progress, etc. as needed; the database 532 is used to store input data obtained by the input unit, learning data used for model learning, and the learning history of the model. The input data stored in the database 532 can be the unprocessed input data itself, or it can be processed data suitable for model learning.

[0031] The learning processor 54 uses training data or training sets to train the artificial neural network. The learning processor 54 learns the artificial neural network by directly acquiring preprocessed data of input data acquired by the processor 55 through the input unit 52, or by acquiring preprocessed input data stored in the database 532.

[0032] Preferably: the learning processor uses the training data to learn to determine the parameters of the artificial neural network, the learning processor 54 may include a memory integrated or implemented in the neural network learning device 5, the learning processor 54 may be implemented using the memory 53, and the learning processor 54 may be implemented using a memory associated with the terminal, such as an external memory directly coupled to the terminal or a memory maintained in a server that communicates with the terminal. Example

[0033] In this embodiment, the temperature value of a specific time and a local area of ​​the pipeline obtained by the pipeline temperature sensor is used as input data to predict the outlet temperature, and the difference between the deep learning predicted value and the actual value is calculated, divided by the actual value, and the average value is taken. Using an expression expressed in percentage, it is judged that the closer the accuracy of the outlet temperature is to 0%, the better the temperature prediction model is.

[0034] like Fig.10 As shown, two representative time series deep learning models, recurrent neural network (RNN) and long short-term memory (LSTM), are applied to accurately determine the characteristics of temperature changes over time.

[0035] The RNN model consists of a chain structure model that repeatedly executes a single layer as time series data is continuously updated; the LSTM model is a structure that adds units, input gates, output gates, and forget gates to the basic RNN structure, which can help solve the problem of skew disappearance in the RNN structure.

[0036] The mean percentage error (MPE) is used to evaluate the performance of the control system based on the deep learning model. The MPE evaluation index is calculated by calculating the actual value of the following formula: With the predicted value The absolute value of the difference between them, taking the average and applying the percentage, if the value is 0, it means that the predicted value measured by the deep learning model is exactly the same as the predicted value measured in the experiment. Where N is the number of samples and t is the true value.

[0037] The devices and components in the present invention include processors, controllers, arithmetic logic units (ALUs), digital signal processors (DSPs), microcomputers, field programmable gate arrays (FPGAs), programmable logic units (PLUs), which can be implemented using one or more general-purpose computers or special-purpose computers, such as microprocessors or any other device capable of executing and responding to instructions. The processing device can execute an operating system (OS) and one or more software applications executed on the operating system. In addition, the processing device can access, store, operate, process and generate data in response to the execution of the software. The processing device may include multiple processing elements and / or multiple types of processing elements. For example, the processing device may include multiple processors or a processor and a controller.

[0038] The software part includes a combination of one or more of computer programs, codes, and instructions. The software part of the present invention is implemented in the form of program commands executed by computer means and can be recorded on a computer-readable medium. The computer-readable medium can include program commands, data files, data structures, etc., individually or in combination. The program commands recorded on the medium can be specially designed and configured for the embodiment.

[0039] The present invention is provided with a heat exchanger module for heating or cooling the medicine supplied from the medicine tank; a control module for performing cooling or heating of the heat exchanger module; a temperature prediction module based on deep learning for predicting the discharge temperature of the medicine discharged from the heat exchanger module; in order to increase the heat transfer area with the heat sink, the medicine pipe is arranged in six straight-through forms inside the first and second heat sinks along the length direction of the heat sink, and a water cooling block is provided between the heat sinks, and a cooling water channel is provided inside the water cooling block. Moreover, since the thermoelectric element is installed between the heat sink and the water cooling block, the heat generated by the thermoelectric element can be cooled to prevent damage caused by thermal shock. Moreover, the medicine pipe is made of fluororesin material with strong chemical resistance and corrosion resistance to avoid reaction with the medicine. The present invention can monitor and control the temperature in the medicine pipe in real time through the pipeline temperature sensor, improve the efficiency of medicine transfer, and accurately control the temperature of the medicine, so as to keep the reaction process of the medicine going smoothly.

[0040] The contents not described in detail in the present invention are all prior art.

[0041] The above description is only a preferred embodiment of the present invention, and is not limited to the description in the specification and implementation. Therefore, any equivalent changes or modifications made according to the structure, features and principles described in the scope of the present invention should be included in the scope of the present invention.

Claims

1. A semiconductor liquid precise temperature control system based on deep learning, comprising a heat exchanger module (1) connected to a chemical liquid pipe, and a neural network learning device (5), characterized in that: The heat exchanger module (1) and the control module (2) are interconnected, the control module (2) and the artificial intelligence module (3) are interconnected, and the neural network learning device (5) is connected to the artificial intelligence module (3): The heat exchanger module (1) comprises a plurality of unit heat exchangers (101), a first heat sink (102), a second heat sink (103), and a water cooling block (104); each of the unit heat exchangers (101) comprises a pipe temperature measuring sensor (10) for measuring the temperature of a medicine liquid pipe at a specified position; a first heat sink (102) and a second heat sink (103) are respectively disposed on both sides of the heat exchanger module (1); the first heat sink (102) and the second heat sink (103) are formed into a square plate having a pipe with a diameter greater than the outer diameter of the medicine liquid pipe; the inlet end (6) of the medicine liquid pipe is disposed on the first heat sink ( 102), the outlet end (7) is arranged in the second heat dissipation block (103), a water cooling block (104) is further arranged between the first heat dissipation block (102) and the second heat dissipation block (103), a cooling water channel is arranged inside the water cooling block (104), a thermoelectric element (105) is arranged between the first heat dissipation block (102) and the water cooling block (104) and between the second heat dissipation block (103) and the water cooling block (104), the thermoelectric element (105) is composed of a plurality of thermocouples, the thermocouples heat or cool the liquid medicine tube at a specified position, and a thermoelectric element controller (1051) controls the direction of the current flowing through the thermocouples; A plurality of T-shaped polytetrafluoroethylene connectors (8) are provided on the medicine liquid pipe between the first heat sink (102) and the second heat sink (103); a plurality of pipeline temperature measurement sensors (10) are provided around the T-shaped polytetrafluoroethylene connector (8); and the T-shaped polytetrafluoroethylene connector (8) is connected to an internal temperature probe (9) for measuring the temperature of medicine in the medicine liquid pipe.

2. According to claim 1, a semiconductor liquid precise temperature control system based on deep learning is characterized in that: The heat exchanger module (1) cools or heats the temperature of the medicine in the medicine pipe, and then sends the temperature measured by the pipeline temperature measurement sensor (10) to the control module (2) through the temperature value sending device (11); the artificial intelligence module (3) learns according to the training data to determine the temperature control model, and the control module (2) predicts the temperature of the medicine to be supplied to the medicine tank according to the temperature control model provided by the artificial intelligence module (3), and performs cooling or heating control according to the predicted temperature.

3. According to claim 1, a semiconductor liquid precise temperature control system based on deep learning is characterized in that: The first heat dissipation block (102) and the second heat dissipation block (103) are respectively provided with six straight-through parts A inside along the length direction, and a plurality of folded parts B outside, and the straight-through parts A are connected to each other in the vertical and horizontal directions through the folded parts B.

4. According to claim 1, a semiconductor liquid precise temperature control system based on deep learning is characterized in that: The water cooling block (104) has an outer shape of a square plate corresponding to the first heat dissipation block (102) and the second heat dissipation block (103).

5. According to claim 1, a semiconductor liquid precise temperature control system based on deep learning is characterized in that: The neural network learning device (5) is a server of a control system, and is based on a deep learning algorithm. The neural network learning device (5) can send a learned facial generation model to a facial generation unit periodically or upon request, or update an installed facial generation model. The neural network learning device (5) comprises a communication unit (51), an input unit (52), a memory (53), a learning processor (54), and a processor (55).

6. The semiconductor liquid precise temperature control system based on deep learning according to claim 5, characterized in that: The communication unit (51) comprises a wireless communication unit and an interface unit, and the communication unit (51) is connected to the drone device through wired or wireless communication and sends or receives data thereto; The input unit (52) acquires training data for model learning and uses the training model to obtain output input data; The memory (53) stores a model learned by a learning processor or a neural network learning device, and the memory includes a model memory (531) or a database (532), wherein the model memory (531) stores a neural network model that is being learned or has been learned by the learning processor; the database (532) is used to store input data obtained by an input unit, learning data used for model learning, and a learning history of the model; The learning processor (54) uses training data or training sets to train the artificial neural network. The learning processor (54) learns the artificial neural network by directly acquiring pre-processed data of input data acquired by the processor (55) through the input unit (52), or by acquiring pre-processed input data stored in a database (532).

7. The semiconductor liquid precise temperature control system based on deep learning according to claim 5, characterized in that: The deep learning algorithm uses a recurrent neural network (RNN) and a long short-term memory (LTSM).

8. The semiconductor liquid precise temperature control system based on deep learning according to claim 1, characterized in that: The pipeline temperature measurement sensor (10) is an RTD temperature sensor.

9. A semiconductor liquid precise temperature control system based on deep learning according to any one of claims 1 to 4, characterized in that: The liquid medicine pipe is made of fluororesin materials with strong chemical resistance and corrosion resistance, including PEA and polytetrafluoroethylene.

Images