WO2020238882A1 - 人工智能辅助印刷电子技术自引导优化提升方法 - Google Patents

人工智能辅助印刷电子技术自引导优化提升方法 Download PDF

Info

Publication number
WO2020238882A1
WO2020238882A1 PCT/CN2020/092253 CN2020092253W WO2020238882A1 WO 2020238882 A1 WO2020238882 A1 WO 2020238882A1 CN 2020092253 W CN2020092253 W CN 2020092253W WO 2020238882 A1 WO2020238882 A1 WO 2020238882A1
Authority
WO
WIPO (PCT)
Prior art keywords
printing
parameters
nozzle
group
pattern
Prior art date
Application number
PCT/CN2020/092253
Other languages
English (en)
French (fr)
Inventor
黄维
王学文
王大朋
李玥
Original Assignee
西北工业大学
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 西北工业大学 filed Critical 西北工业大学
Priority to US16/970,873 priority Critical patent/US11882664B2/en
Publication of WO2020238882A1 publication Critical patent/WO2020238882A1/zh

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B41—PRINTING; LINING MACHINES; TYPEWRITERS; STAMPS
    • B41J—TYPEWRITERS; SELECTIVE PRINTING MECHANISMS, i.e. MECHANISMS PRINTING OTHERWISE THAN FROM A FORME; CORRECTION OF TYPOGRAPHICAL ERRORS
    • B41J29/00—Details of, or accessories for, typewriters or selective printing mechanisms not otherwise provided for
    • B41J29/38—Drives, motors, controls or automatic cut-off devices for the entire printing mechanism
    • B41J29/393—Devices for controlling or analysing the entire machine ; Controlling or analysing mechanical parameters involving printing of test patterns
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B41—PRINTING; LINING MACHINES; TYPEWRITERS; STAMPS
    • B41J—TYPEWRITERS; SELECTIVE PRINTING MECHANISMS, i.e. MECHANISMS PRINTING OTHERWISE THAN FROM A FORME; CORRECTION OF TYPOGRAPHICAL ERRORS
    • B41J2/00—Typewriters or selective printing mechanisms characterised by the printing or marking process for which they are designed
    • B41J2/005—Typewriters or selective printing mechanisms characterised by the printing or marking process for which they are designed characterised by bringing liquid or particles selectively into contact with a printing material
    • B41J2/01—Ink jet
    • B41J2/21—Ink jet for multi-colour printing
    • B41J2/2132—Print quality control characterised by dot disposition, e.g. for reducing white stripes or banding
    • B41J2/2135—Alignment of dots
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/12—Digital output to print unit, e.g. line printer, chain printer
    • G06F3/1201—Dedicated interfaces to print systems
    • G06F3/1202—Dedicated interfaces to print systems specifically adapted to achieve a particular effect
    • G06F3/1203—Improving or facilitating administration, e.g. print management
    • G06F3/1208—Improving or facilitating administration, e.g. print management resulting in improved quality of the output result, e.g. print layout, colours, workflows, print preview
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/12—Digital output to print unit, e.g. line printer, chain printer
    • G06F3/1293—Printer information exchange with computer
    • G06F3/1294—Status or feedback related to information exchange
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/12—Digital output to print unit, e.g. line printer, chain printer
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • H—ELECTRICITY
    • H05—ELECTRIC TECHNIQUES NOT OTHERWISE PROVIDED FOR
    • H05K—PRINTED CIRCUITS; CASINGS OR CONSTRUCTIONAL DETAILS OF ELECTRIC APPARATUS; MANUFACTURE OF ASSEMBLAGES OF ELECTRICAL COMPONENTS
    • H05K3/00—Apparatus or processes for manufacturing printed circuits
    • H05K3/10—Apparatus or processes for manufacturing printed circuits in which conductive material is applied to the insulating support in such a manner as to form the desired conductive pattern
    • H05K3/12—Apparatus or processes for manufacturing printed circuits in which conductive material is applied to the insulating support in such a manner as to form the desired conductive pattern using thick film techniques, e.g. printing techniques to apply the conductive material or similar techniques for applying conductive paste or ink patterns
    • H05K3/1241—Apparatus or processes for manufacturing printed circuits in which conductive material is applied to the insulating support in such a manner as to form the desired conductive pattern using thick film techniques, e.g. printing techniques to apply the conductive material or similar techniques for applying conductive paste or ink patterns by ink-jet printing or drawing by dispensing
    • H05K3/125—Apparatus or processes for manufacturing printed circuits in which conductive material is applied to the insulating support in such a manner as to form the desired conductive pattern using thick film techniques, e.g. printing techniques to apply the conductive material or similar techniques for applying conductive paste or ink patterns by ink-jet printing or drawing by dispensing by ink-jet printing

Definitions

  • the invention relates to the field of printed electronics and computing science and technology, and particularly relates to an artificial intelligence-assisted self-guided optimization and promotion method of printed electronic technology.
  • the method described in the literature only analyzes the influence of the piezoelectric waveform on the printing effect, ignoring the influence of other factors on the printing effect, and the number of experimental groups is small, only the best parameters in the experimental group can be obtained, and the actual situation cannot be obtained.
  • the best printing parameters under. It takes a lot of time to analyze many factors that affect the printing effect by this method, and it is difficult to obtain the best printing parameters.
  • the purpose of the present invention is to provide an artificial intelligence-assisted self-guided optimization and improvement method of printed electronics technology, which combines machine learning technology and printed electronics technology to determine the optimal printing parameters according to the factors that affect the printing effect, and reduces the number of printers.
  • the user's early time spent testing the printing effect is good for practicality.
  • the present invention provides an artificial intelligence-assisted self-guided optimization and improvement method of printed electronic technology, including:
  • Step 1 Influencing factors of print quality and experimental group settings
  • Step 2 Design the printing pattern
  • the line width of the straight wires in the printing pattern are all 10 ⁇ m
  • the line spacing of the straight wires in the printing pattern are all 110 ⁇ m
  • the line width of the curves in the printing pattern are all 80 ⁇ m.
  • the line spacing of the curves in the printed pattern are all 160 ⁇ m;
  • Step 3 Print out sample patterns respectively according to the 24 printing parameter combinations
  • the linear position and the curved position of the sample pattern were characterized by an optical microscope. As the number of nozzle holes increases, the details of the printed pattern become worse and worse, and the wires are connected together; the distance between the nozzle substrates is large, and the wires are bent; The smaller the average value of the width and the closer to the design value, and the smaller the standard deviation of the line width, the better the printing effect; the data characterizing the printing effect includes 24 printing parameter combinations, the average value of the line width of each sample pattern and each The standard deviation of the line width of a combined sample pattern;
  • Step 5 Machine learning technology analyzes the data
  • Step 6 The printing parameters are sent back to the user's computer to guide the user to optimize the printing quality
  • the computer sends the printing parameters obtained in step 5 back to the printer control program.
  • the control program automatically modifies the printer printing parameters and prints to obtain an optimized printing pattern; the optimized printing pattern is characterized under an optical microscope and the line width
  • the average and standard deviation of the data are uploaded to the computer and calculated by machine learning to optimize the printing effect.
  • the six variables of the number of nozzle holes of the printer, the number of repeated printing, the printing speed, the temperature of the printing substrate, the distance between the substrate of the print head and the ink jet strength of the print head are divided into six groups, and each group is composed of four uniformly changing parameters.
  • 24 combinations of printing parameters including:
  • the number of repeated printing is determined as the second group, and the parameters of the second group are 1, 2, 4, and 6 times;
  • the parameters of the third group are respectively 50mm/s, 100mm/s, 150mm/s and 200mm/s;
  • the ink jet strength of the nozzle is determined as a sixth group, and the parameters of the sixth group are 65%, 75%, 85%, and 95%, respectively.
  • setting the other five groups of parameter conditions as fixed printing parameters specifically includes:
  • the number of nozzle holes of the printer are printed according to the first set of parameters.
  • the other five sets of parameters are as follows: the number of repeated printing is 1 time, the printing speed is 150mm/s, and the printing The substrate temperature is room temperature, the nozzle substrate spacing is 0.1mm, and the nozzle ink jet strength is 100%;
  • the repeated printing times are printed according to the second set of parameters.
  • the other five sets of parameters are as follows: the number of printer nozzles is 1, the printing speed is 150mm/s, and the printing The substrate temperature is room temperature, the nozzle substrate spacing is 0.1mm, and the nozzle ink jet strength is 100%;
  • the printing speed is printed according to the third group of parameters.
  • the other five groups of parameters are as follows: the number of printer nozzles is 1, the number of repeated printing is 1, and the temperature of the printed substrate At room temperature, the nozzle substrate spacing is 0.1mm, and the nozzle ink jet strength is 100%;
  • the printing substrate temperature is printed according to the fourth group of parameters.
  • the other five groups of parameters are as follows: the number of printer nozzles is 1, the number of repeated printing is 1, and the printing The speed is 150mm/s, the print head substrate spacing is 0.1mm, and the print head ink jet strength is 100%;
  • the print head substrate spacing is printed according to the fifth set of parameters.
  • the other five sets of parameters are as follows: the number of printer nozzles is one, and the number of repeated printing is one. The speed is 150mm/s, the temperature of the printing substrate is room temperature, and the ink jet strength of the nozzle is 100%;
  • the inkjet strength of the nozzle is printed according to the sixth group of parameters.
  • the other five groups of parameters are as follows: the number of printer nozzles is 1 and the number of repeated printing is 1 , The printing speed is 150mm/s, the printing substrate temperature is room temperature, and the nozzle substrate spacing is 0.1mm.
  • the characterizing the linear position and the curved position of the sample pattern through an optical microscope specifically includes:
  • the five positions are three different positions of a straight wire and two different positions of a curve;
  • the 24 printing parameter combinations, the average of the line width of the sample pattern of each combination and the standard deviation of the line width of the sample pattern of each combination were used as the data to characterize the printing effect.
  • the analysis of data through machine learning technology to obtain the printing parameters corresponding to the best printing effect specifically includes:
  • the present invention discloses the following technical effects:
  • the present invention proposes an artificial intelligence-assisted printed electronic technology self-guided optimization and promotion method.
  • the method combines machine learning technology and printed electronic technology.
  • the user sets experimental groups according to variables that affect the printing quality of microelectronic printers.
  • the medium parameter uses a microelectronic printer to print samples to characterize the printing effect and judge the quality of the printing.
  • the characterization results are analyzed through machine learning to obtain the printing parameters corresponding to the best printing effect, and these parameters are fed back to the user.
  • the user sets the printer according to the feedback parameters to improve the printing quality.
  • the invention only needs to set up a few simple sets of experiments according to several factors affecting the printing effect to obtain the optimal printing parameters, reduces the time that the printer user spends in the early stage testing the printing effect, and has good practicability.
  • Figure 1 is a circuit diagram of a printed pattern designed by an embodiment of the present invention
  • Figure 2 is a circuit diagram of a sample pattern according to an embodiment of the present invention.
  • Fig. 3 is a schematic diagram of sampling points in an embodiment of the present invention.
  • 4A is an optical microscope image of the first position of the straight line according to the embodiment of the present invention.
  • 4B is an optical microscope image of the second position of the straight line according to the embodiment of the present invention.
  • 4C is an optical microscope image of the third position of the straight line according to the embodiment of the present invention.
  • 4D is an optical microscope image of the first position of the curve of the embodiment of the present invention.
  • 4E is an optical microscope image of the second position of the curve of the embodiment of the present invention.
  • 5A is an optical microscope image of a printed sample when the influencing factor of print quality is the number of nozzle holes according to an embodiment of the present invention
  • FIG. 5B is an optical microscope image of the printed sample when the printing quality influence factor is the number of repeated printing according to the embodiment of the present invention.
  • FIG. 5C is an optical microscope image of the printed sample when the printing quality influencing factor is the printing speed according to the embodiment of the present invention.
  • 5D is an optical microscope image of the printed sample when the printing quality influencing factor is the temperature of the printing substrate according to the embodiment of the present invention.
  • FIG. 5E is an optical microscope image of the printed sample when the printing quality influencing factor is the distance between the nozzle substrates according to the embodiment of the present invention.
  • Fig. 5F is an optical microscope image of a printed sample when the printing quality influencing factor is the ink jet strength of the nozzle according to the embodiment of the present invention.
  • 6A is a schematic diagram of the influence on the printing effect when the printing quality influencing factor is the number of nozzle holes according to the embodiment of the present invention.
  • 6B is a schematic diagram of the influence on the printing effect when the printing quality influencing factor is the number of repeated printing according to the embodiment of the present invention.
  • 6C is a schematic diagram of the influence of the printing quality on the printing effect when the printing speed is the printing speed according to the embodiment of the present invention.
  • 6D is a schematic diagram of the influence of the printing quality on the printing effect when the temperature of the printing substrate is the influence factor of printing quality according to the embodiment of the present invention
  • 6E is a schematic diagram of the influence on the printing effect when the printing quality influencing factor is the distance between the nozzle substrates according to the embodiment of the present invention.
  • FIG. 6F is a schematic diagram of the influence of the printing quality on the printing effect when the ink jet strength of the nozzle according to the embodiment of the present invention.
  • the purpose of the present invention is to provide an artificial intelligence-assisted self-guided optimization and improvement method for printed electronic technology, which combines machine learning technology with printed electronic technology, determines the optimal printing parameters according to the factors that affect the printing effect, and reduces the printer user's initial cost It is practical when testing the printing effect.
  • the invention provides an artificial intelligence-assisted printing electronic technology self-guided optimization and promotion method.
  • the printing effect is evaluated and quantified by the user controlling the printing parameters, the quantified data is uploaded to the computer, and the data is machine-learned to analyze the specific parameters
  • the printing effect is the best, and the result is sent back to the interface of the printer control software used by the user, and the user can perform printing operations based on this data.
  • Step 1 Influencing factors of print quality and experimental group settings.
  • the printer used in the present invention is a new type of microelectronic printer.
  • the factors that may affect the printing quality of the printer are analyzed, and the number of printer nozzles, the number of repeated printing, printing speed, printing substrate temperature,
  • the six variables, the distance between the print head substrate and the ink jet strength of the print head, are the main factors affecting the print quality.
  • the factors that affect the printing effect also include the voltage waveform that controls the ink jet of the printer head, the air pressure between the ink jet cartridge and the ink bag, and the air pressure in the laboratory where the printer is located. Divide these 6 variables into 6 groups, and each group consists of 4 uniformly changing parameters. The parameters of each group can also be composed of five parameters or more.
  • the factors that affect print quality and print parameters are shown in Table 1.
  • the other five groups of conditions were set as: repeat printing once, printing speed 150mm /s, the temperature of the printing substrate is room temperature, the distance between the nozzle and the substrate is 0.1mm, and the inkjet strength of the nozzle is 100%; when studying the influence of the number of repeated printing on the print quality, the number of repeated printing is set to 1, 2, 4, and 6 respectively And print separately, the other five groups of conditions are set as: the number of printer nozzles is 1, the printing speed is 150mm/s, the printing substrate temperature is room temperature, the distance between the nozzle and the substrate is 0.1mm, and the nozzle inkjet strength is 100%; the printing speed is being studied When it affects the print quality, set the printing speed to 50, 100, 150, 200mm/s and print separately.
  • the other five groups of conditions are set as follows: the number of printer nozzles is 1, the number of repeated printing is 1 time, and the substrate is printed The temperature is room temperature, the distance between the nozzle and the substrate is 0.1mm, and the inkjet strength of the nozzle is 100%; when studying the influence of the printing substrate temperature on the print quality, the printing substrate temperature is set to 21°C, 30°C, 40°C, and 50°C respectively.
  • the other five groups of conditions are set as: the number of printer nozzles is 1, the number of repetitive printing is 1 time, the printing speed is 150mm/s, the distance between the nozzle and the substrate is 0.1mm, and the inkjet strength of the nozzle is 100%; the distance between the nozzle and the substrate is being studied When the print quality is affected, the distance between the nozzle and the substrate is set to 0.1, 0.6, 1.1, 2.1mm and print respectively.
  • the other five groups of conditions are set as: the number of printer nozzles is 1, the number of repetitive printing is 1 time, and the printing The speed is 150mm/s, the temperature of the printing substrate is room temperature, and the inkjet strength of the nozzle is 100%; when studying the influence of the inkjet strength of the nozzle on the print quality, the inkjet strength of the nozzle is set to 65%, 75%, and 85% of the total voltage. , 95% and print separately, the other five groups of conditions are set as: the number of printer nozzles is 1, the number of repeated printing is 1 time, the printing speed is 150mm/s, the printing substrate temperature is room temperature, and the distance between the nozzle and the substrate is 0.1mm. There are 16 nozzles integrated on one nozzle.
  • Step two design the print pattern.
  • the printing pattern needs to be analyzed to analyze the influence of the printing parameters on the straight line and the curve. Therefore, the straight line and the curve structure are designed. As shown in Figure 1, the line width of the straight wire of the printed pattern is 10 ⁇ m, and the line spacing Both are 110 ⁇ m. The line width of the printed pattern curve is 80 ⁇ m, and the line spacing is 160 ⁇ m.
  • Step three Print out sample patterns separately from 24 kinds of printing parameter combinations.
  • the printing parameters according to the combination of 24 sets of printing parameters, and print out the physical image of the designed printing pattern.
  • the physical image is a sample pattern. As shown in Figure 2, the printed physical image has the same appearance as the designed printed pattern, and the lines are clear and identifiable.
  • Figure 4A is the first straight line
  • Fig. 4B is the optical microscope image of the second position of the line
  • Fig. 4C is the optical microscope image of the third position of the line
  • Fig. 4D is the optical microscope image of the first position of the curve
  • Fig. 4E It is the optical microscope image of the first position of the curve.
  • Figures 4A-4E are the optical microscope images of 5 positions under a set of printing parameters.
  • Figure 5A is the optical microscope image of the printed sample when the print quality influencing factor is the number of nozzles
  • Figure 5B is the optical microscope image of the printed sample when the print quality influencing factor is the number of repeated printing
  • Figure 5C is when the print quality influencing factor is the printing speed
  • the optical microscope image of the printed sample Figure 5D is the optical microscope image of the printed sample when the printing quality influencing factor is the temperature of the printed substrate
  • Figure 5E is the optical microscope image of the printed sample when the printing quality influencing factor is the print head substrate spacing
  • Figure 5F is the printing quality
  • the influencing factor is the optical microscope image of the printed sample when the ink jet strength of the nozzle is ejected.
  • Figures 5A-5F are the optical microscope images of the printed sample under all 24 parameter combinations.
  • the ink jet strength of the nozzle is controlled by the ink jet voltage. The lower the voltage, the less ink will be ejected. There may even be no ink jet.
  • the optical microscope image shows that some wires are disconnected. These three factors directly determine the quality of the printed product. The increase in the number of printing results in darker color of the wire, which is reflected in the electrical performance that the impedance of the wire is smaller and the performance is better.
  • the abscissa of Fig. 6A represents the number of nozzle holes; the abscissa of Fig. 6B represents the number of prints; the abscissa of Fig. 6C represents the printing speed, in mm/s; the abscissa of Fig. 6D represents the temperature of the printed substrate, in °C; the abscissa of Fig. 6E represents The pitch of the jetted substrate, the unit is mm; the abscissa of Figure 6F represents the ratio of the maximum voltage of the voltage waveform that controls the ejection of the nozzle to the maximum allowable voltage waveform of the printer's ejection voltage waveform during the printing operation.
  • Printing speed is the speed at which the print head moves. For the smallest print head substrate spacing condition, the print speed has little effect on the printing effect, and the line width changes little with the increase of the printing speed. As the temperature increases, the line width increases slightly, but the standard deviation value is more stable.
  • Step 5 Machine learning technology analyzes the data.
  • Model selection consists of three basic parts: data collection, data preprocessing, and model evaluation.
  • Data preprocessing includes two steps: random oversampling and standardization.
  • a random oversampling method is used to randomly select a small number of labeled samples of a certain type, and then put them back into the data set until the data sets of the two classes have the same Count. Then standardize the balanced data set and scale the feature values to speed up the convergence of modeling training.
  • a nested cross-validation scheme is adopted.
  • Model evaluation mainly includes two steps: 1) inner cross-validation: select the optimal hyperparameters for each candidate model; 2) outer cross-validation: test each model to evaluate the performance of each model on a new data set . Then select the best model based on the generated test results.
  • the invention uses the GBDT method to perform machine learning to obtain the optimal printing parameters.
  • Incremental adaptive model (incremental learning): first list the existing laboratory-printed data combinations, and then perform experimental tests on the combinations, generate data samples as the initial training set, and determine that each class has at least 10 samples to satisfy The minimum sample criterion for training the model. Using the GDBT training method for these samples, first perform hierarchical cross-validation on the initial training set and select the optimal hyperparameters of the model; then use the trained model to predict the probability that the remaining parameter combination is "good”. Then, the parameter combination with the highest probability is tested experimentally, the label is generated, and it is included in the training set of the next experiment.
  • Gradient descent tree Gradient descent tree (GBDT) is a typical type of gradient enhancement.
  • F m (x) is the decision function
  • ⁇ m is the learning parameter
  • m is the label of the basis decision maker
  • M is the total number of basis decision makers.
  • ROC curve is a curve that measures the performance of a two-class model.
  • the GBDT method is adopted. This method combines parameter adjustment and model selection. The selected model has a small error on the training set and the test set.
  • the outer cross-validation the data set is layered to achieve optimal parameter training, and the probability value of the predicted sample is given; in the inner cross-validation, cross-validation is implemented in the training set to adjust the parameters, and then the results are fed back to
  • the hyperparameters are tuned, and the updated hyperparameters are used to continue training the model.
  • TPR and NPR The true positive rate (TPR) is defined as the correctly predicted positive rate divided by the total number of true positive samples.
  • the true negative rate (NPR) represents the number of negative correct predictive values divided by the total number of true negative samples.
  • the positive type is expressed as "good effect", and the negative type is expressed as "bad effect”.
  • the prediction result comes from the model generated by the data set. Try to ensure the balance between the number of positive and negative samples to make the ROC curve of the model more balanced.
  • Machine learning is performed on the average and standard deviation data obtained under each group of printing conditions in the above-mentioned manner.
  • the printing effect is approximately linearly changed with the printing parameters, that is, when other conditions remain unchanged, as the 4 parameters under a certain printing condition gradually increase (decrease), the printing effect will gradually become better or worse, and it is approximately linear
  • Step 6 The parameters are returned to the user's computer to guide the user to optimize the print quality.
  • the computer transmits this printing parameter back to the printer control program, and the control program automatically modifies the printing parameters of the microelectronic printer and performs printing, and then the optimized printing pattern can be obtained. Characterize the newly obtained printed pattern under an optical microscope, and upload the average and standard deviation data of the line width to the computer. With the increase of subsequent users, the sample data will be more and more, and the database will be gradually expanded. The result will be more and more accurate, and the printing effect will be more satisfactory.
  • the invention provides an artificial intelligence-assisted self-guided optimization and promotion method for printed electronics technology, which combines machine learning technology and printed electronics technology.
  • the user sets experimental groups according to variables that affect the printing quality of microelectronic printers, and according to the parameters in the experimental group Use a microelectronic printer to print samples to characterize the printing effect and judge the quality of the printing.
  • the characterization results are analyzed through machine learning to obtain the printing parameters corresponding to the best printing effect, and these parameters are fed back to the user.
  • the user sets the printer according to the feedback parameters to improve the printing quality.
  • the invention only needs to set up a few simple sets of experiments according to several factors affecting the printing effect to obtain the optimal printing parameters, reduces the time that the printer user spends in the early stage testing the printing effect, and has good practicability.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Quality & Reliability (AREA)
  • Manufacturing & Machinery (AREA)
  • Microelectronics & Electronic Packaging (AREA)
  • Ink Jet (AREA)
  • Accessory Devices And Overall Control Thereof (AREA)

Abstract

一种人工智能辅助印刷电子技术自引导优化提升方法。该方法将机器学习技术与印刷电子技术相结合，用户根据影响微电子打印机打印质量的变量，设置实验组别，根据实验组中参数使用微电子打印机打印样品，对打印效果进行表征，评判打印质量好坏；通过机器学习的方式对表征结果进行分析，得到最好的打印效果对应的打印参数，并将这些参数反馈给用户，用户根据反馈参数设置打印机，提高打印质量。该方法只需根据影响打印效果的几个因素即可得到最优的打印参数，减少了打印机用户前期花费在测试打印效果的时间。

Description

人工智能辅助印刷电子技术自引导优化提升方法
本申请要求于2019年5月31日提交中国专利局、申请号为201910470758.4、发明名称为“人工智能辅助印刷电子技术自引导优化提升方法”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本发明涉及印刷电子和计算科学技术领域，特别是涉及一种人工智能辅助印刷电子技术自引导优化提升方法。
背景技术
文献“压电波形对喷墨打印电极的调控规律，发光学报，2017，Vol.38，No.5，P617-P622”公开了一种通过改变压电波形来研究其对喷墨打印效果影响规律的方法。该方法在原有压电波形的基础上，改变压电波形的加压速率和脉冲持续时间，使用改变后的压电波形分别进行打印操作，表征打印效果，分析得到最好的打印效果对应的压电波形。文献所述方法只分析了压电波形对打印效果的影响，忽略了其它多种因素对打印效果的影响，而且实验组数较少，只能得到实验组中的最好参数，无法得到实际情况下的最好打印参数。通过该方法对影响打印效果的众多因素一一进行分析会花费大量的时间，且难以得到最好的打印参数。
发明内容
基于此，本发明的目的是提供一种人工智能辅助印刷电子技术自引导优化提升方法，将机器学习技术与印刷电子技术相结合，根据影响打印效果的因素确定最优的打印参数，减少了打印机用户前期花费在测试打印效果的时间，实用性好。
为实现上述目的，本发明提供了一种人工智能辅助印刷电子技术自引导优化提升方法，包括：
步骤一、打印质量影响因素及实验组设置；
确认打印机、打印墨水和打印基底符合要求，将打印机喷孔数量、重复打印次数、打印速度、打印基板温度、喷头基板间距和喷头喷墨力度六个变量分成六组，每组由四个均匀变化的参数组成，共有24种打印参数组合；所述六个变量为影响打印质量的因素；
按所述六组中任一组参数条件打印时，将其它五组参数条件设定为固定的打印参数；
步骤二、设计打印图案；
确定好打印参数后，设计打印图案，所述打印图案中直导线的线宽均为10μm，所述打印图案中直导线的线距均为110μm，所述打印图案中曲线的线宽均为80μm，所述打印图案中曲线的线距均为160μm；
步骤三、根据所述24种打印参数组合分别打印出样品图案；
根据所述24组打印参数组合设置打印参数，将设计的打印图案打印出实物图；所述实物图为样品图案；
步骤四、打印效果表征；
通过光学显微镜对所述样品图案的直线位置和曲线位置进行表征，随着喷孔数量增多，打印的图案细节越来越差，导线连接在一起；喷头基板间距大，导线的弯曲程度大；线宽的平均值越小且越接近设计值、线宽的标准差越小则打印效果越好；打印效果表征的数据包括24种打印参数组合、每一组合的样品图案线宽的平均值和每一组合的样品图案线宽的标准差；
步骤五、机器学习技术分析数据；
将打印效果表征的数据上传到计算机，通过机器学习技术分析数据，得到最好的打印效果对应的打印参数；
步骤六、打印参数回传用户电脑，引导用户优化打印质量；
计算机将步骤五得到的打印参数回传到打印机控制程序，控制程序自动修改打印机打印参数并进行打印，得到优化后的打印图案；将优化后的打印图案在光学显微镜下进行表征，并将线宽的平均值和标准差数据上传计算机，通过机器学习计算，优化打印效果。
可选的，所述将打印机喷孔数量、重复打印次数、打印速度、打印基板温度、喷头基板间距和喷头喷墨力度六个变量分成六组，每组由四个均匀变化的参数组成，共有24种打印参数组合，具体包括：
将打印机喷孔数量确定为第一组，所述第一组的参数分别为1个、2个、4个和6个；
将重复打印次数确定为第二组，所述第二组的参数分别为1次、2次、 4次和6次；
将打印速度确定为第三组，所述第三组的参数分别为50mm/s、100mm/s、150mm/s和200mm/s；
将打印基板温度确定为第四组，所述第四组的参数分别为21℃、30℃、40℃和50℃；
将喷头基板间距确定为第五组，所述第五组的参数分别为0.1mm、0.6mm、1.1mm和2.1mm；
将喷头喷墨力度确定为第六组，所述第六组的参数分别为65％、75％、85％和95％。
可选的，所述按所述六组中任一组参数条件打印时，将其它五组参数条件设定为固定的打印参数，具体包括：
在确定喷孔数量对打印质量的影响时，打印机喷孔数量按照所述第一组的参数分别进行打印，其它五组参数条件为：重复打印次数为1次，打印速度为150mm/s，打印基板温度为室温，喷头基板间距为0.1mm，喷头喷墨力度为100％；
在确定重复打印次数对打印质量的影响时，重复打印次数按照所述第二组的参数分别进行打印，其它五组参数条件为：打印机喷孔数量为1个，打印速度为150mm/s，打印基板温度为室温，喷头基板间距为0.1mm，喷头喷墨力度为100％；
在确定打印速度对打印质量的影响时，打印速度按照所述第三组的参数分别进行打印，其它五组参数条件为：打印机喷孔数量为1个，重复打印次数为1次，打印基板温度为室温，喷头基板间距为0.1mm，喷头喷墨力度为100％；
在确定打印基板温度对打印质量的影响时，打印基板温度按照所述第四组的参数分别进行打印，其它五组参数条件为：打印机喷孔数量为1个，重复打印次数为1次，打印速度为150mm/s，喷头基板间距为0.1mm，喷头喷墨力度为100％；
在确定喷头基板间距对打印质量的影响时，喷头基板间距按照所述第五组的参数分别进行打印，其它五组参数条件为：打印机喷孔数量为1个，重复打印次数为1次，打印速度为150mm/s，打印基板温度为室温， 喷头喷墨力度为100％；
在确定喷头喷墨力度对打印质量的影响时，喷头喷墨力度按照所述第六组的参数分别进行打印，其它五组参数条件为：打印机喷孔数量为1个，重复打印次数为1次，打印速度为150mm/s，打印基板温度为室温，喷头基板间距为0.1mm。
可选的，所述通过光学显微镜对所述样品图案的直线位置和曲线位置进行表征，具体包括：
在所述打印图案中选取五个位置，所述五个位置分别为直导线的三个不同位置和曲线的两个不同位置；
通过光学显微镜观察所述样品图案，得到每一种打印参数组合在所述五个位置的光学显微镜图像；
通过所述光学显微镜图像测量线宽，每个位置处测量三条线的线宽，得到直导线的九个测量值和曲线的六个测量值；
确定所述直导线的九个测量值的平均值和标准差，以及确定所述曲线的六个测量值的平均值和标准差；
将24种打印参数组合、每一组合的样品图案线宽的平均值和每一组合的样品图案线宽的标准差作为打印效果表征的数据。
可选的，所述通过机器学习技术分析数据，得到最好的打印效果对应的打印参数，具体包括：
根据所述打印效果表征的数据采用GBDT算法确定最好的打印效果对应的打印参数；所述最好的打印效果对应的打印参数为在打印参数集合中样品图案线宽的平均值最小、标准差最小时的打印参数；所述打印参数集合中打印参数的个数为a b,a表示均匀变化的参数，b表示影响打印质量的因素，a＝4，b＝6。
根据本发明提供的具体实施例，本发明公开了以下技术效果：
本发明提出了一种人工智能辅助印刷电子技术自引导优化提升方法，该方法将机器学习技术与印刷电子技术相结合，用户根据影响微电子打印机打印质量的变量，设置实验组别，根据实验组中参数使用微电子打印机打印样品，对打印效果进行表征，评判打印质量好坏。通过机器学习的方式对表征结果进行分析，得到最好的打印效果对应的打印参数，并将这些 参数反馈给用户，用户根据反馈参数设置打印机，提高打印质量。随着微电子打印机用户的增多，计算机得到的样本数据也会越来越多，计算机机器学习的结果也会更加精确，反馈给用户的打印参数打印出的样品效果更好。本发明只需根据影响打印效果的几个因素设置简单的几组实验即可得到最优的打印参数，减少了打印机用户前期花费在测试打印效果的时间，实用性好。
说明书附图
为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对实施例中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1为本发明实施例设计的打印图案电路图；
图2为本发明实施例样品图案电路图；
图3为本发明实施例取样点示意图；
图4A为本发明实施例直线的第1个位置的光学显微镜图像；
图4B为本发明实施例直线的第2个位置的光学显微镜图像；
图4C为本发明实施例直线的第3个位置的光学显微镜图像；
图4D为本发明实施例曲线的第1个位置的光学显微镜图像；
图4E为本发明实施例曲线的第2个位置的光学显微镜图像；
图5A为本发明实施例打印质量影响因素为喷孔个数时打印样品的光学显微镜图像；
图5B为本发明实施例打印质量影响因素为重复打印次数时打印样品的光学显微镜图像；
图5C为本发明实施例打印质量影响因素为打印速度时打印样品的光学显微镜图像；
图5D为本发明实施例打印质量影响因素为打印基板温度时打印样品的光学显微镜图像；
图5E为本发明实施例打印质量影响因素为喷头基板间距时打印样品的光学显微镜图像；
图5F为本发明实施例打印质量影响因素为喷头喷墨力度时打印样品 的光学显微镜图像；
图6A为本发明实施例打印质量影响因素为喷孔个数时对打印效果的影响示意图；
图6B为本发明实施例打印质量影响因素为重复打印次数时对打印效果的影响示意图；
图6C为本发明实施例打印质量影响因素为打印速度时对打印效果的影响示意图；
图6D为本发明实施例打印质量影响因素为打印基板温度时对打印效果的影响示意图；
图6E为本发明实施例打印质量影响因素为喷头基板间距时对打印效果的影响示意图；
图6F为本发明实施例打印质量影响因素为喷头喷墨力度时对打印效果的影响示意图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
本发明的目的是提供一种人工智能辅助印刷电子技术自引导优化提升方法，将机器学习技术与印刷电子技术相结合，根据影响打印效果的因素确定最优的打印参数，减少了打印机用户前期花费在测试打印效果的时间，实用性好。
为使本发明的上述目的、特征和优点能够更加明显易懂，下面结合附图和具体实施方式对本发明作进一步详细的说明。
本发明提供的一种人工智能辅助印刷电子技术自引导优化提升方法，通过用户控制打印参数对打印效果进行评估量化，将量化后的数据上传计算机，对数据进行机器学习，分析出具体哪种参数下的打印效果最好，将这一结果回传到用户使用的打印机控制软件界面，用户能够根据这一数据进行打印操作。
方法具体步骤如下：
步骤一、打印质量影响因素及实验组设置。
本发明中使用的打印机是新型微电子打印机，首先在打印墨水和打印基底确认合适的情况下分析可能影响打印机打印质量的因素，确定打印机喷孔数量、重复打印次数、打印速度、打印基板温度、喷头基板间距和喷头喷墨力度这6个变量是影响打印质量的主要因素。影响打印效果的因素还包括控制打印机喷头喷墨的电压波形、喷墨盒与墨囊间气压大小、打印机所在实验室的气压大小等。将这6个变量分成6组，每组都由4个均匀变化的参数组成。每组的参数也可以由五个参数或者更多参数组成。影响打印质量的因素与打印参数如表1所示。
表1 影响打印质量的因素与打印参数
  一 二 三 四
打印机喷孔数量(个) 1 2 4 6
重复打印次数(次) 1 2 4 6
打印速度(mm/s) 50 100 150 200
打印基板温度℃ 21 30 40 50
喷头与基板间距mm 0.1 0.6 1.1 2.1
喷头喷墨力度 65％ 75％ 85％ 95％
根据表1可知，一共有4x4x4x4x4x4＝4096种打印参数的组合，单一用户很难把所有参数的组合都打印出来，所以采取分组实验的方式，即在研究某组条件时，将其它五组条件设定为固定的打印参数，这一固定参数对实验组打印效果的影响很小。在研究喷孔数量对打印质量的影响时，打印机喷孔数量分别设定为1，2，4，6个并分别进行打印，其它五组条件设定为：重复打印次数1次，打印速度150mm/s，打印基板温度为室温，喷头与基板间距0.1mm，喷头喷墨力度100％；在研究重复打印次数对打印质量的影响时，重复打印次数分别设定为1，2，4，6次并分别进行打印，其它五组条件设定为：打印机喷孔数量1个，打印速度150mm/s，打印基板温度为室温，喷头与基板间距0.1mm，喷头喷墨力度100％；在研究打印速度对打印质量的影响时，打印速度分别设定为50，100，150，200mm/s并分别进行打印，其它五组条件设定为：打印机喷孔数量1个，重复打印次数1次，打印基板温度为室温，喷头与基板间距0.1mm，喷 头喷墨力度100％；在研究打印基板温度对打印质量的影响时，打印基板温度分别设定为21℃，30℃，40℃，50℃并分别进行打印，其它五组条件设定为：打印机喷孔数量1个，重复打印次数1次，打印速度150mm/s，喷头与基板间距0.1mm，喷头喷墨力度100％；在研究喷头与基板间距对打印质量的影响时，喷头与基板间距分别设定为0.1，0.6，1.1，2.1mm并分别进行打印，其它五组条件设定为：打印机喷孔数量1个，重复打印次数1次，打印速度150mm/s，打印基板温度为室温，喷头喷墨力度100％；在研究喷头喷墨力度对打印质量的影响时，喷头喷墨力度分别设定为总电压的65％，75％，85％，95％并分别进行打印，其它五组条件设定为：打印机喷孔数量1个，重复打印次数1次，打印速度150mm/s，打印基板温度为室温，喷头与基板间距0.1mm。一个喷头上集成有16个喷孔。
步骤二、设计打印图案。
确定好打印参数后，设计打印图案，需要分析打印参数对直线和曲线分别有什么影响，因此设计了直线和曲线结构，如图1所示，打印图案直导线的线宽均为10μm，线距均为110μm。打印图案曲线线宽为80μm，线距为160μm。
步骤三、将24种打印参数组合分别打印出样品图案。
根据24组打印参数组合设置打印参数，将设计的打印图案打印出实物图，实物图为样品图案，如图2所示，打印出的实物图与设计的打印图案外观一致，线条清晰可辨。
步骤四、打印效果表征。
如图3所示，在设计的打印图案上选定直线的3个不同位置和曲线的2个不同位置，通过光学显微镜分别对样品的直线和曲线位置进行表征，图4A是直线的第1个位置的光学显微镜图像，图4B是直线的第2个位置的光学显微镜图像，图4C是直线的第3个位置的光学显微镜图像，图4D是曲线的第1个位置的光学显微镜图像，图4E是曲线的第1个位置的光学显微镜图像，图4A-图4E是一组打印参数下的5个位置的光学显微镜图像。
图5A是打印质量影响因素为喷孔个数时打印样品的光学显微镜图像，图5B是打印质量影响因素为重复打印次数时打印样品的光学显微镜 图像，图5C是打印质量影响因素为打印速度时打印样品的光学显微镜图像，图5D是打印质量影响因素为打印基板温度时打印样品的光学显微镜图像，图5E是打印质量影响因素为喷头基板间距时打印样品的光学显微镜图像，图5F是打印质量影响因素为喷头喷墨力度时打印样品的光学显微镜图像，图5A-图5F是所有的24组参数组合下的打印样品的光学显微镜图像。可见，喷孔数量、喷头基底间距、喷头喷墨力度这3个因素对打印效果影响较大。可明显观察到随着喷孔数量增多，打印的图案细节越来越差，导线连接在一起。喷头基底间距较大，墨水从喷头喷出到基底的路径变长，受到喷头水平移动的影响，墨滴做抛物线运动落到基底上，间距越大墨滴落在基底偏差越大，间距越大，导线的弯曲程度越大(直导线打印后出现弯曲现象较为明显，曲线也有这种弯曲现象，但是不明显)。喷头喷墨力度由喷墨的电压控制，电压越低喷出的墨量越少，甚至可能出现不喷墨的情况，光学显微镜图像可见部分导线出现断开的情况。这3个因素直接决定打印产品的好坏。打印次数的增多导致导线的颜色更深，反映到电性能上就是导线的阻抗更小，性能更好。
通过图5A-图5F的光学显微镜图像测量线宽，每个位置处测量3条线的线宽，这样每个打印图案在直线部分会有9个测量值，在曲线部分有6个测量值，分别对直线部分的9个测量值和曲线部分的6个测量值取平均值和标准差值，以平均值和标准差值作为评判打印效果好坏的标准，平均值越小且越接近设计值、标准差越小则打印效果越好。将这些平均值和标准差数据做成表格观察它们的变化趋势，图6A-图6F分别展示了6个因素对打印效果的影响情况，图6A-图6F中的纵坐标均表示导线的宽度，图6A横坐标表示喷孔个数；图6B横坐标表示打印次数；图6C横坐标表示打印速度，单位是mm/s；图6D横坐标表示打印基板温度，单位是℃；图6E横坐标表示喷空基底间距，单位是mm；图6F横坐标表示打印操作时控制喷头喷墨的电压波形的最大电压与打印机喷墨电压波形允许的最大电压占比。随着打印次数的增加直线和曲线的线宽都逐渐增加，而且线宽的标准差也逐渐趋于稳定。打印速度即喷头移动的速度，对于最小的喷头基底间距条件而言，打印速度对打印效果影响很小，线宽随着打印速度的增加变化很小。随着温度的升高线宽略有增加，但标准差值更加稳定。
步骤五、机器学习技术分析数据。
将24种打印参数组合及对应打印参数组合下样品图案的线宽的平均值和标准差上传到计算机，通过机器学习技术编写程序，为了实现有机导电墨水电路打印参数的优化，需要确定各种打印参数之间的模糊关系。因此，有必要通过研究已有的打印数据，选择一个最能揭示这种模糊关系的机器学习(Machine Learning，ML)模型，生成最优的打印条件，并获得较高的成功率。
模型选择：模型选择由三个基本部分组成:数据采集、数据预处理、模型评估。数据预处理包括随机过采样和标准化两个步骤。为了解决数据集中样本分布不平衡的问题，采用随机过采样的方法，通过反复随机选择数量较少的某一类被标记样本，然后将其放回数据集中，直到两个类的数据集具有相同的计数。然后标准化已经平衡的数据集，将特征数值进行缩放以加快建模训练的收敛。为了解决小数据集问题，采用了嵌套交叉验证方案。模型评价主要包括两个步骤:1)内层交叉验证：为每个候选模型选择最优超参数；2)外层交叉验证：测试每个模型，以评估每个模型在新数据集上的性能。然后根据产生的测试结果选择最佳模型。本发明采用GBDT方法进行机器学习得到最优打印参数。
增量自适应模型(增量学习)：首先列出已有的实验室打印的数据组合，然后对组合进行实验测试，生成数据样本作为初始训练集，确定每个类至少有10个样本，满足训练模型的最少样本准则。对这些样本采用GDBT训练方法，首先对初始训练集进行分层交叉验证，选择模型的最优超参数；然后利用训练后的模型预测剩余参数组合“效果好”的概率。然后对具有最高概率的参数组合进行实验测试，生成标签，并将其包含到下一个试验的训练集中。同样的步骤不断重复，此过程达到临界点时停止(此时所有剩余集中的组合都被预测为“效果不好”，即最高预测概率低于50.0％)，此时停止训练，因为模型对其余组合的信任度较低。
梯度下降树(GBDT)：梯度下降树(GBDT)是一种典型的梯度增强类型，梯度提升树用的是梯度下降法，梯度下降树(GBDT)通过M个基决策器h m(m＝1,2,...,M)产生的一组值进行决策：
Figure PCTCN2020092253-appb-000001
式中，F m(x)为决策函数，γ m为学习参数，m为基决策器标号，M为基决策器的总数。
已知N个训练数据
Figure PCTCN2020092253-appb-000002
和可微分的损失函数f L(y,F(x))，F(x)表示拟合函数值，y表示实验真值，损失函数表示拟合函数值和实验真值的差距，利用迭代的方式进行训练：F m(x)＝F m-1(x)+γ mh m(x)，其中γ m是通过最小化下一个模型对应的损失函数而确定的。在每一步中，残差γ m是本轮模型F m-1(x)对应的损失函数的负梯度。之后，利用
Figure PCTCN2020092253-appb-000003
训练h m(x)。x i为第i个数据，r mi为第i个数据的残差。
ROC曲线(接受者操作特性曲线，receiver operating characteristic curve)：ROC曲线是衡量二分类模型性能的曲线。为了绘制出模型的ROC曲线图，采用GBDT方法，此方法将调参和模型挑选结合起来，选择的模型在训练集和测试集上的误差很小。在外层交叉验证中，将数据集进行分层实现最优参数训练，并且给出需要预测样本的概率值；在内层交叉验证中，在训练集中实现交叉验证实现参数调整，然后将结果反馈到超参数调优机制中，对超参数调优，并使用更新后的超参数继续训练模型。
TPR与NPR:真阳性率(TPR)定义为正确预测阳性率除以真实阳性样本总数。真负率(NPR)表示正确预测值为负的个数除以真阴性样本总数。正类表示为“效果好”，负类表示为“效果不好”。预测结果来自于数据集生成的模型。尽量保证正类和负类样本数目的平衡，可以使得模型的ROC曲线更加平衡。
对各组打印条件下得到的平均值和标准差值数据通过上述方式进行机器学习。打印效果是随着打印参数近似线性变化的，即其它条件不变时，随着某一打印条件下4个参数逐渐增加(减少)，打印效果也会逐渐变好或变坏，且近似呈线性变化，这样就可以通过机器学习技术分析出6个打印条件、4个打印参数变量下共4 6＝4096种打印参数组合下的打印样品线宽均值最小，标准差最小时的打印参数。
步骤六、参数回传用户电脑，引导用户优化打印质量。
计算机将这一打印参数回传到打印机控制程序，控制程序自动修改微电子打印机打印参数并进行打印，即可得到优化后的打印图案。将新得到的打印图案在光学显微镜下进行表征，并将线宽的平均值和标准差数据上 传计算机，随着后续用户增多，样本数据也会越来越多，逐步扩充数据库，机器学习计算的结果也会越来越精确，打印效果也更加满足要求。
本发明提供的一种人工智能辅助印刷电子技术自引导优化提升方法，将机器学习技术与印刷电子技术相结合，用户根据影响微电子打印机打印质量的变量，设置实验组别，根据实验组中参数使用微电子打印机打印样品，对打印效果进行表征，评判打印质量好坏。通过机器学习的方式对表征结果进行分析，得到最好的打印效果对应的打印参数，并将这些参数反馈给用户，用户根据反馈参数设置打印机，提高打印质量。随着微电子打印机用户的增多，计算机得到的样本数据也会越来越多，计算机机器学习的结果也会更加精确，反馈给用户的打印参数打印出的样品效果更好。本发明只需根据影响打印效果的几个因素设置简单的几组实验即可得到最优的打印参数，减少了打印机用户前期花费在测试打印效果的时间，实用性好。
本文中应用了具体个例对本发明的原理及实施方式进行了阐述，以上实施例的说明只是用于帮助理解本发明的方法及其核心思想；同时，对于本领域的一般技术人员，依据本发明的思想，在具体实施方式及应用范围上均会有改变之处。综上所述，本说明书内容不应理解为对本发明的限制。

Claims (5)

  1. 一种人工智能辅助印刷电子技术自引导优化提升方法，其特征在于，所述方法包括：
    步骤一、打印质量影响因素及实验组设置；
    确认打印机、打印墨水和打印基底符合要求，将打印机喷孔数量、重复打印次数、打印速度、打印基板温度、喷头基板间距和喷头喷墨力度六个变量分成六组，每组由四个均匀变化的参数组成，共有24种打印参数组合；所述六个变量为影响打印质量的因素；
    按所述六组中任一组参数条件打印时，将其它五组参数条件设定为固定的打印参数；
    步骤二、设计打印图案；
    确定好打印参数后，设计打印图案，所述打印图案中直导线的线宽均为10μm，所述打印图案中直导线的线距均为110μm，所述打印图案中曲线的线宽均为80μm，所述打印图案中曲线的线距均为160μm；
    步骤三、根据所述24种打印参数组合分别打印出样品图案；
    根据所述24组打印参数组合设置打印参数，将设计的打印图案打印出实物图；所述实物图为样品图案；
    步骤四、打印效果表征；
    通过光学显微镜对所述样品图案的直线位置和曲线位置进行表征，随着喷孔数量增多，打印的图案细节越来越差，导线连接在一起；喷头基板间距大，导线的弯曲程度大；线宽的平均值越小且越接近设计值、线宽的标准差越小则打印效果越好；打印效果表征的数据包括24种打印参数组合、每一组合的样品图案线宽的平均值和每一组合的样品图案线宽的标准差；
    步骤五、机器学习技术分析数据；
    将打印效果表征的数据上传到计算机，通过机器学习技术分析数据，得到最好的打印效果对应的打印参数；
    步骤六、打印参数回传用户电脑，引导用户优化打印质量；
    计算机将步骤五得到的打印参数回传到打印机控制程序，控制程序自动修改打印机打印参数并进行打印，得到优化后的打印图案；将优化后的打印图案在光学显微镜下进行表征，并将线宽的平均值和标准差数据上传 计算机，通过机器学习计算，优化打印效果。
  2. 根据权利要求1所述的人工智能辅助印刷电子技术自引导优化提升方法，其特征在于，所述将打印机喷孔数量、重复打印次数、打印速度、打印基板温度、喷头基板间距和喷头喷墨力度六个变量分成六组，每组由四个均匀变化的参数组成，共有24种打印参数组合，具体包括：
    将打印机喷孔数量确定为第一组，所述第一组的参数分别为1个、2个、4个和6个；
    将重复打印次数确定为第二组，所述第二组的参数分别为1次、2次、4次和6次；
    将打印速度确定为第三组，所述第三组的参数分别为50mm/s、100mm/s、150mm/s和200mm/s；
    将打印基板温度确定为第四组，所述第四组的参数分别为21℃、30℃、40℃和50℃；
    将喷头基板间距确定为第五组，所述第五组的参数分别为0.1mm、0.6mm、1.1mm和2.1mm；
    将喷头喷墨力度确定为第六组，所述第六组的参数分别为65％、75％、85％和95％。
  3. 根据权利要求2所述的人工智能辅助印刷电子技术自引导优化提升方法，其特征在于，所述按所述六组中任一组参数条件打印时，将其它五组参数条件设定为固定的打印参数，具体包括：
    在确定喷孔数量对打印质量的影响时，打印机喷孔数量按照所述第一组的参数分别进行打印，其它五组参数条件为：重复打印次数为1次，打印速度为150mm/s，打印基板温度为室温，喷头基板间距为0.1mm，喷头喷墨力度为100％；
    在确定重复打印次数对打印质量的影响时，重复打印次数按照所述第二组的参数分别进行打印，其它五组参数条件为：打印机喷孔数量为1个，打印速度为150mm/s，打印基板温度为室温，喷头基板间距为0.1mm，喷头喷墨力度为100％；
    在确定打印速度对打印质量的影响时，打印速度按照所述第三组的参数分别进行打印，其它五组参数条件为：打印机喷孔数量为1个，重复打 印次数为1次，打印基板温度为室温，喷头基板间距为0.1mm，喷头喷墨力度为100％；
    在确定打印基板温度对打印质量的影响时，打印基板温度按照所述第四组的参数分别进行打印，其它五组参数条件为：打印机喷孔数量为1个，重复打印次数为1次，打印速度为150mm/s，喷头基板间距为0.1mm，喷头喷墨力度为100％；
    在确定喷头基板间距对打印质量的影响时，喷头基板间距按照所述第五组的参数分别进行打印，其它五组参数条件为：打印机喷孔数量为1个，重复打印次数为1次，打印速度为150mm/s，打印基板温度为室温，喷头喷墨力度为100％；
    在确定喷头喷墨力度对打印质量的影响时，喷头喷墨力度按照所述第六组的参数分别进行打印，其它五组参数条件为：打印机喷孔数量为1个，重复打印次数为1次，打印速度为150mm/s，打印基板温度为室温，喷头基板间距为0.1mm。
  4. 根据权利要求3所述的人工智能辅助印刷电子技术自引导优化提升方法，其特征在于，所述通过光学显微镜对所述样品图案的直线位置和曲线位置进行表征，具体包括：
    在所述打印图案中选取五个位置，所述五个位置分别为直导线的三个不同位置和曲线的两个不同位置；
    通过光学显微镜观察所述样品图案，得到每一种打印参数组合在所述五个位置的光学显微镜图像；
    通过所述光学显微镜图像测量线宽，每个位置处测量三条线的线宽，得到直导线的九个测量值和曲线的六个测量值；
    确定所述直导线的九个测量值的平均值和标准差，以及确定所述曲线的六个测量值的平均值和标准差；
    将24种打印参数组合、每一组合的样品图案线宽的平均值和每一组合的样品图案线宽的标准差作为打印效果表征的数据。
  5. 根据权利要求4所述的人工智能辅助印刷电子技术自引导优化提升方法，其特征在于，所述通过机器学习技术分析数据，得到最好的打印效果对应的打印参数，具体包括：
    根据所述打印效果表征的数据采用GBDT算法确定最好的打印效果对应的打印参数；所述最好的打印效果对应的打印参数为在打印参数集合中样品图案线宽的平均值最小、标准差最小时的打印参数；所述打印参数集合中打印参数的个数为a b,a表示均匀变化的参数，b表示影响打印质量的因素，a＝4，b＝6。
PCT/CN2020/092253 2019-05-31 2020-05-26 人工智能辅助印刷电子技术自引导优化提升方法 WO2020238882A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US16/970,873 US11882664B2 (en) 2019-05-31 2020-05-26 Artificial intelligence-assisted printed electronics self-guided optimization method

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910470758.4 2019-05-31
CN201910470758.4A CN110077126B (zh) 2019-05-31 2019-05-31 人工智能辅助印刷电子技术自引导优化提升方法

Publications (1)

Publication Number Publication Date
WO2020238882A1 true WO2020238882A1 (zh) 2020-12-03

Family

ID=67422934

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/092253 WO2020238882A1 (zh) 2019-05-31 2020-05-26 人工智能辅助印刷电子技术自引导优化提升方法

Country Status (3)

Country Link
US (1) US11882664B2 (zh)
CN (1) CN110077126B (zh)
WO (1) WO2020238882A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116306875A (zh) * 2023-05-18 2023-06-23 成都理工大学 基于空间预学习与拟合的排水管网样本增量学习方法
CN117591046A (zh) * 2023-12-26 2024-02-23 广州市森扬电子科技有限公司 一种深度卷积网络打印效果优化的方法、设备及存储介质

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110077126B (zh) 2019-05-31 2020-02-14 西北工业大学 人工智能辅助印刷电子技术自引导优化提升方法
JP2021062565A (ja) * 2019-10-16 2021-04-22 セイコーエプソン株式会社 情報処理装置、学習装置、情報処理装置の制御方法、及び、プログラム
EP3954540A1 (en) * 2020-08-11 2022-02-16 Schott AG Method for printing an image on a substrate and corresponding system
CN114193938A (zh) * 2021-11-29 2022-03-18 长城信息股份有限公司 一种基于激光单元自适应调整的激光打印机成像控制系统
CN115358159B (zh) * 2022-10-24 2023-04-07 浙江大学计算机创新技术研究院 一种多目标3d打印工艺参数优化方法
JP2024077715A (ja) * 2022-11-29 2024-06-10 株式会社Ｓｃｒｅｅｎホールディングス 印刷システム、調整支援方法、および調整支援プログラム

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1439519A (zh) * 2002-02-19 2003-09-03 佳能株式会社 喷墨记录装置及其控制方法和程序
CN101746128A (zh) * 2008-12-19 2010-06-23 佳能株式会社 喷墨打印设备和打印方法
US20110248243A1 (en) * 2009-11-30 2011-10-13 Omega Optics, Inc. Carbon nanotube field effect transistor for printed flexible/rigid electronics
JP2018069684A (ja) * 2016-11-02 2018-05-10 株式会社リコー 情報処理装置、情報処理システム、及び情報処理方法
CN108172524A (zh) * 2017-12-01 2018-06-15 华南理工大学 一种通过压电波形调控tft阵列源漏电极薄膜质量的方法
CN109130550A (zh) * 2018-08-21 2019-01-04 嘉兴学院 一种电流体动力学喷印的人工智能控制系统及控制方法
JP2019022166A (ja) * 2017-07-20 2019-02-07 京セラドキュメントソリューションズ株式会社 画像処理装置、画像処理方法及び画像処理プログラム
CN110077126A (zh) * 2019-05-31 2019-08-02 西北工业大学 人工智能辅助印刷电子技术自引导优化提升方法

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2375365A1 (en) * 1999-05-27 2001-02-15 Patterning Technologies Limited Method of forming a masking pattern on a surface
WO2005027504A1 (ja) * 2003-09-10 2005-03-24 Seiko Epson Corporation 出力画像データ生成装置および出力画像データ生成方法
JP2007245666A (ja) * 2006-03-17 2007-09-27 Sony Corp サーマルヘッド及びプリンタ装置
JP5361172B2 (ja) * 2007-11-16 2013-12-04 キヤノン株式会社 記録制御装置および記録制御方法
CN103241025B (zh) * 2013-04-28 2015-02-04 京东方科技集团股份有限公司 一种有机薄膜的喷墨打印方法
CN108986086A (zh) * 2018-07-05 2018-12-11 福州大学 印刷显示面板喷墨打印像素缺陷检测与分类方法及其装置
EP4094947A3 (en) * 2021-05-27 2023-03-08 The Boeing Company Printing system for generating nozzle firing patterns based on positional offsets

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1439519A (zh) * 2002-02-19 2003-09-03 佳能株式会社 喷墨记录装置及其控制方法和程序
CN101746128A (zh) * 2008-12-19 2010-06-23 佳能株式会社 喷墨打印设备和打印方法
US20110248243A1 (en) * 2009-11-30 2011-10-13 Omega Optics, Inc. Carbon nanotube field effect transistor for printed flexible/rigid electronics
JP2018069684A (ja) * 2016-11-02 2018-05-10 株式会社リコー 情報処理装置、情報処理システム、及び情報処理方法
JP2019022166A (ja) * 2017-07-20 2019-02-07 京セラドキュメントソリューションズ株式会社 画像処理装置、画像処理方法及び画像処理プログラム
CN108172524A (zh) * 2017-12-01 2018-06-15 华南理工大学 一种通过压电波形调控tft阵列源漏电极薄膜质量的方法
CN109130550A (zh) * 2018-08-21 2019-01-04 嘉兴学院 一种电流体动力学喷印的人工智能控制系统及控制方法
CN110077126A (zh) * 2019-05-31 2019-08-02 西北工业大学 人工智能辅助印刷电子技术自引导优化提升方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
NING, HONGLONG ET AL.: "Regulation Rules of Piezoelectric Waveform on Ink-jet Printing Electrode", CHINESE JOURNAL OF LUMINESCENCE, vol. 38,, no. 5, 31 May 2017 (2017-05-31), pages 618 - 622, ISSN: 1000-7032, DOI: 20200703102038Y *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116306875A (zh) * 2023-05-18 2023-06-23 成都理工大学 基于空间预学习与拟合的排水管网样本增量学习方法
CN117591046A (zh) * 2023-12-26 2024-02-23 广州市森扬电子科技有限公司 一种深度卷积网络打印效果优化的方法、设备及存储介质
CN117591046B (zh) * 2023-12-26 2024-07-19 广州市森扬电子科技有限公司 一种深度卷积网络打印效果优化的方法、设备及存储介质

Also Published As

Publication number Publication date
CN110077126B (zh) 2020-02-14
US11882664B2 (en) 2024-01-23
US20220194093A1 (en) 2022-06-23
CN110077126A (zh) 2019-08-02

Similar Documents

Publication Publication Date Title
WO2020238882A1 (zh) 人工智能辅助印刷电子技术自引导优化提升方法
US8941876B2 (en) Ink use amount evaluation apparatus and method, and inkjet apparatus
CN113094980B (zh) 一种基于iga-dnn的焊膏印刷质量预测方法及系统
EP3616931B1 (en) Printhead waveform adjustment
US20210183036A1 (en) Enhanced print defect detection
US20060158478A1 (en) Circuit modeling and selective deposition
WO2024174346A1 (zh) Oled像素层喷印落点位置规划与喷印控制方法
CN110674614B (zh) 一种基于rx mask中心点阵的信号眼图分析方法
JP2022526119A (ja) 塗料、ワニス、印刷インキ、研削用樹脂、色素濃縮物、または他のコーティング剤の組成物の生成方法
CN115371981A (zh) 喷头筛选方法、装置、设备及介质
CN116330845B (zh) 一种新型显示喷印制造液滴体积自适应智能调控方法
US20200269573A1 (en) Medium classification mechanism
US20230035323A1 (en) Inkjet printing method and apparatus
US20220032616A1 (en) Driving waveform determining method, non-transitory computer-readable storage medium storing driving waveform determining program, liquid ejecting apparatus, and driving waveform determining system
CN115008900A (zh) 一种柔性显示喷印薄膜边缘直线度控制方法和系统
CN113492589A (zh) 具有可变模型的补偿特性曲线（dumc）
US20230095675A1 (en) Controlling precision systems using free topology waveforms
CN111216466B (zh) 喷墨打印的方法、装置以及计算机设备
JP2024077715A (ja) 印刷システム、調整支援方法、および調整支援プログラム
KR20240095465A (ko) 열가소성 화합물에 대한 레시피를 생성하는 시스템 및 방법
CN117309343A (zh) 喷头筛选方法、装置、设备及介质
CN114817002A (zh) 基于反馈机制的自然语言处理模型测试用例约简方法
CN116383725A (zh) 一种喷墨打印用阵列化喷孔的智能筛选方法及装置
CN118003769A (zh) 一种喷墨打印巨量喷嘴的液滴体积均值调控方法
CN118685848A (zh) 一种电镀加工参数智能预测方法及系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20815027

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20815027

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 20815027

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 24/05/2022)