目的 针对介电层制备过程中厚度控制和均匀性优化的关键问题,建立压电阀喷墨打印工艺参数与介电层打印厚度和均匀性的关系模型,进而完成工艺参数的寻优,实现对介电层厚度和均匀性的调控。方法 运用传统拟合方法（响应面法（RSM））与智能算法（反向传播神经网络（BP）、基于粒子群算法的BP神经网络（PSO-BP）、基于改进粒子群算法的BP神经网络（IPSO-BP）等）,构建了延迟时间、打印速度、线间距、打印层数与介电层厚度及均匀性的定量关系模型,并通过方差分析（ANOVA）探讨各参数的显著性及交互作用机理,比较各模型的预测精度。选择介电层打印的目标厚度为200、400、600、800 μm,以打印厚度的误差最小化及均匀性最优控制为优化目标,通过RSM及IPSO算法对工艺参数进行反函数寻优,并对比2种方法的优化效果。结果 IPSO-BP神经网络的预测能力优于其他模型,该模型预测的介电层厚度和均匀性的均方根误差（10.935 7、2.457 4）和平均绝对误差（4.283 4、1.170 9）最小,决定系数（99.89%、95.49%）最高;采用IPSO算法寻优得到的工艺参数打印的介电层厚度和均匀性的平均相对误差为3.50%和16.75%,优于RSM算法的平均相对误差15.21%和29.06%。结论 相较于SVM、RF、BP、PSO-BP神经网络,IPSO-BP神经网络在处理本研究中复杂非线性问题时适应性更好、全局搜索能力更强、预测精度更高,在压电阀打印介电层的过程中,IPSO-BP-IPSO方法能够通过优化工艺参数实现介电层的厚度控制和均匀性优化。
Abstract
To address the critical issues of thickness control and uniformity optimization in dielectric layer fabrication, the work aims to establish relationship models between the piezoelectric valve inkjet printing process parameters and the printed thickness and uniformity of dielectric layers to complete parameter optimization, thereby achieving precise control of both thickness and uniformity in dielectric layer formation. The quantitative relationship model between the process parameters (delay time, printing speed, line spacing, and number of printed layers) and the thickness and uniformity of the dielectric layer was established with Response Surface Methodology (RSM) and intelligent algorithms (Backpropagation neural network (BP), Particle Swarm-optimized BP neural network (PSO-BP) and Improved Particle Swarm-optimized BP neural network (IPSO-BP)). Then, the Analysis of Variance (ANOVA) was applied to assess the statistical significance and interaction mechanisms of each parameter, and the prediction accuracy of various models was compared. The target thicknesses selected for dielectric layer printing were 200, 400, 600, and 800 μm. With the minimization of printing thickness error and the optimal control of uniformity as optimization objectives, the Response Surface Methodology (RSM) and Improved Particle Swarm Optimization (IPSO) algorithm were selected to solve the optimization problems to obtain the optimized process parameters and the optimization results were compared. The prediction accuracy of the IPSO-BP neural network surpassed that of all other models and achieved the lowest root mean square errors (10.935 7 for thickness and 2.457 4 for uniformity) and mean absolute errors (4.283 4 for thickness and 1.170 9 for uniformity), with the highest coefficients of determination (99.89% for thickness and 95.49% for uniformity). The mean relative errors of thickness and uniformity of the printed dielectric layer under process parameters obtained by the IPSO algorithm were 3.50% and 16.75%, much lower than those obtained by the RSM method (15.21% and 29.06%). Compared to SVM, RF, BP and PSO-BP neural networks, the IPSO-BP algorithm has better adaptability, stronger global search ability and higher prediction accuracy in dealing with the complex nonlinear problems in this study. In the process of printing dielectric layers with piezoelectric valves, the IPSO-BP-IPSO method can achieve precise control of dielectric layer thickness and the optimization of its uniformity by optimizing process parameters.
关键词
压电阀打印 /
介电层 /
响应面法 /
BP神经网络 /
改进粒子群算法
Key words
piezoelectric valve printing /
dielectric layer /
response surface methodology /
BP neural network /
improved particle swarm optimization
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] SMITHA P S, BABU V S, ASMIN M K.Investigation on Barrier Layers of PT/BaTiO3/PT Based Thin Film Capacitors[J]. ChemNanoMat, 2024, 10(12): e20240-0316.
[2] CHEN M, PENG B Y, GUO X Y, et al.Polyethylene Interfacial Dielectric Layer for Organic Semiconductor Single Crystal Based Field-Effect Transistors[J]. Chinese Chemical Letters, 2024, 35(4): 109051.
[3] LIU R L, JI B, LEI M, et al.Triple-Gradient Based Dielectric Layer for Flexible Capacitive Sensor with Broad Sensing Linearity and High Sensitivity[J]. Applied Materials Today, 2025, 42: 102614.
[4] HEATH J P, HARDING J H, SINCLAIR D C, et al.Electric Field Enhancement in Ceramic Capacitors Due to Interface Amplitude Roughness[J]. Journal of the European Ceramic Society, 2019, 39(4): 1170-1177.
[5] WANG L Y, SUN G D, YUAN S G.Chemical Vapor Deposition Growth of 2D Ferroelectric Materials for Device Applications[J]. Advanced Materials Technologies, 2024, 9(9): 2301973.
[6] HUANG M L, WU Y Z, YI N B, et al.Structural Colouration on Textile Fabrics with Thin-Film Coating via Magnetron Sputtering: A Review[J]. Surface Engineering, 2022, 38(10/11/12): 830-845.
[7] YASEEN M, KHATTAK M A K, KHAN A, et al. State-of-the-Art Electrochromic Thin Films Devices, Fabrication Techniques and Applications: A Review[J]. Nanocomposites, 2024, 10(1): 1-40.
[8] LATIF M, JIANG Y, KIM J.Additively Manufactured Flexible Piezoelectric Lead Zirconate Titanate-Nanoce- llulose Films with Outstanding Mechanical Strength, Dielectric and Piezoelectric Properties[J]. Materials To- day Advances, 2024, 21: 100478.
[9] WANG D X, JIANG W, LI S R, et al.A Comprehensive Review on Combinatorial Film via High-Throughput Techniques[J]. Materials, 2023, 16(20): 6696.
[10] CAROU-SENRA P, RODRÍGUEZ-POMBO L, AWAD A, et al. Inkjet Printing of Pharmaceuticals[J]. Advanced Materials, 2024, 36(11): 2309164.
[11] KWON K S, RAHMAN M K, PHUNG T H, et al.Review of Digital Printing Technologies for Electronic Materials[J]. Flexible and Printed Electronics, 2020, 5(4): 043003.
[12] MCGHEE J R, MIDDLEMISS R B, SOUTHEE D J, et al.Flexible, all Metal-Oxide Capacitors for Printed Electronics[C]// 2018 IEEE 13th Nanotechnology Materials and Devices Conference (NMDC). Portland, OR, USA. IEEE, 2018: 1-4.
[13] KILISZKIEWICZ M, PRZYBYLSKI D, FELBA J, et al.Structural Characterization of Inkjet Printed Capacitor Layers in Various Technological Conditions[J]. Soldering & Surface Mount Technology, 2020, 32(4): 235-240.
[14] VASIMALLA S, SUBBARAO N V V, GEDDA M, et al. Effects of Dielectric Material, HMDS Layer, and Channel Length on the Performance of the Perylenediimide-Based Organic Field-Effect Transistors[J]. ACS Omega, 2017, 2(6): 2552-2560.
[15] JIANG H W, QIAN R Z, YANG T H, et al.Inkjet-Printed Dielectric Layer for the Enhancement of Electrowetting Display Devices[J]. Nanomaterials, 2024, 14(4): 347.
[16] 焦晓阳, 刘建芳, 谷峰春, 等. 压电喷射点胶阀的喷射性能分析及实验研究[J]. 四川大学学报(工程科学版), 2013, 45(2): 193-198.
JIAO X Y, LIU J F, GU F C, et al.Analysis and Research on Jetting Ability of the Jet Dispensing Valve Driven by a Piezostack[J]. Journal of Sichuan University (Engineering Science Edition), 2013, 45(2): 193-198.
[17] 邓珺珺, 邓圭玲, 彭雯, 等. 压电驱动喷射点胶阀系统性能的仿真与实验[J]. 传感器与微系统, 2023, 42(1): 46-49.
DENG J J, DENG G L, PENG W, et al.Simulation and Experiment of System Performance of Piezostack-Driven Jetting Dispenser[J]. Transducer and Microsystem Technologies, 2023, 42(1): 46-49.
[18] ZHANG C, TAO M, LING M X.Numerical Investigation of Highly Viscous Droplet Generation Based on Level Set Method[J]. Physica Scripta, 2023, 98(11): 115007.
[19] CHOI J E, SONG J, LEE Y H, et al.Deep Neural Network Modeling of Multiple Oxide/Nitride Deposited Dielectric Films for 3D-NAND Flash[J]. Applied Science and Convergence Technology, 2020, 29(6): 190-194.
[20] SIM H E, LEE M U, HONG S J.Virtual Metrology of Multiple Dielectric Layer Thickness for 3D-NAND Deposition Process[J]. IEEE Transactions on Semiconductor Manufacturing, 2025, 38(2): 240-250.
[21] ZHANG T P, JI P F, TIAN D Y, et al.Prediction of COD Degradation in Fenton Oxidation Treatment of Kitchen Anaerobic Wastewater Based on IPSO-BP Neural Network[J]. Journal of Electrical and Computer Engineering, 2025, 2025(1): 3213686.
[22] CAO Y W, MA R, ZHAO K J, et al.Predicting Peak Particle Velocity in Pre-Splitting of Gas-Producing Devices Using Improved Particle Swarm Optimization Algorithm[J]. Scientific Reports, 2025, 15: 13663.
[23] 李莉, 张赛, 何强, 等. 响应面法在试验设计与优化中的应用[J]. 实验室研究与探索, 2015, 34(8): 41-45.
LI L, ZHANG S, HE Q, et al.Application of Response Surface Methodology in Experiment Design and Optimization[J]. Research and Exploration in Laboratory, 2015, 34(8): 41-45.
[24] 范勇, 裴勇, 杨广栋, 等. 基于改进PSO-BP神经网络的爆破振动速度峰值预测[J]. 振动与冲击, 2022, 41(16): 194-203.
FAN Y, PEI Y, YANG G D, et al.Prediction of Blasting Vibration Velocity Peak Based on an Improved PSO-BP Neural Network[J]. Journal of Vibration and Shock, 2022, 41(16): 194-203.
[25] 朱必武, 蒋昊, 刘筱, 等. 基于改进PSO-BP神经网络预测中高应变速率轧制AZ31镁合金板的抗拉强度[J]. 中国有色金属学报, 2025, 57(4): 1-16.
ZHU B W, JIANG H, LIU X, et al.Prediction of the Tensile Strength of AZ31 Magnesium Alloy Sheet Rolled at Medium-High Strain Rate based on Improved PSO-BP Neural Network[J]. The Chinese Journal of NonferrousMetals, 2025, 57(4): 1-16.
[26] 郑伯文, 邓芝超, 罗振豪, 等. 基于改进粒子群优化神经网络的拖拉机排放方法研究[J]. 农业机械学报, 2023, 54(S2): 417-426.
ZHENG B W, DENG Z C, LUO Z H, et al.Research on Tractor Emission Method Based on Improved Particle Swarm Optimization Neural Network[J]. Transactions of the Chinese Society for Agricultural Machinery, 2023, 54(S2): 417-426.
基金
国家重点研发计划（2022YFB4602800）; 西安市科技计划重点产业链关键核心技术攻关项目（23LLRH0079）
PDF(7138 KB)
渝公网安备 50010702501719号 渝ICP备15012534号-4
