稳健估计用于氮化铝金属化工艺的研究

应用化学镀镍的方法实现了氮化铝的金属化。为得到较大的氮化铝金属化层粘附力，运用基于稳健估计的神经网络研究氮化铝金属化中化学镀镍的反应参数与金属层粘附力的关系。为使神经网络更加稳健，本文根据统计学原理，在前馈神经网络基础上，采取稳健估计方法改进神经网络。建立了定量预测粘附力性能的模型，并进行实验验证。确定金属化工艺中稳定的优化工作区域。结果表明，稳健估计方法既有传统神经网络的优点，又有较强的抵抗异常     

稳健估计用于氮化铝金属化工艺的研究

应用化学镀镍的方法实现了氮化铝的金属化。为得到较大的氮化铝金属化层粘附力 ,运用基于稳健估计的神经网络研究氮化铝金属化中化学镀镍的反应参数与金属层粘附力的关系。为使神经网络更加稳健 ,本文根据统计学原理 ,在前馈神经网络基础上 ,采取稳健估计方法改进神经网络。建立了定量预测粘附力性能的模型 ,并进行实验验证。确定金属化工艺中稳定的优化工作区域。结果表明 ,稳健估计方法既有传统神经网络的优点 ,又有较强的抵抗异常值的能力 ,具有较广泛的实用性。     

基于粗大误差估计的表面测量稳健滤波方法研究

将粗大误差估计理论与高斯滤波方法相结合,提出了一种新的表面测量稳健方法。首先由kσ准则对表面轮廓中的畸形信号进行判定、识别、修正,详细论述了相应算法;然后由移动平均逼近方法实现高斯滤波,建立表面轮廓中线;最后由原始轮廓和得到的稳健滤波中线分离出表面粗糙度轮廓,实现了稳健滤波目的。该方法算法简单、易于实现,对表面畸形信号能够较好地抑制。     

含铝超高碳钢等温球化工艺的研究

在对超高碳钢进行铝合金化的基础上,利用成分不均匀奥氏体化加热控制,提出了锻态超高碳钢(UHCs-1.6Al)有效的无形变球化工艺,并利用扫描电镜对先共析碳化物的析出及球化进行了观察,分析了碳化物的球化机理.结果表明,通过对UHCs-1.6Al成分不均匀奥氏体化加热,使先共析碳化物在随后冷却时以粒状形式在基体上弥散析出,利用等温过程使其球化长大,并使奥氏体继续冷却时发生离异共析转变,从而获得球化组织;碳化物颗粒的尺寸可以通过等温温度和等温时间控制;当奥氏体化进行充分时,先共析碳化物析出的孕育期延长,部分碳化物在随后共析转变中以片状形式形成;UHCs-1.6Al最佳球化工艺的奥氏体化透烧温度和等温温度分别为850～870℃和780～800℃.     

Surface Preparation of Aluminum Nitride for Metallization

Currently, the surface preparation of aluminum nitride (AlN) substrates prior to metallization includes an aqueous cleaning step. Surface reactions that occur in this step cause performance and reliability issues with AlN substrates to be used in microelectronic packaging. There is a lack of published data on the reactivity of AlN substrates with common solvents. This study investigated the effects of different solvents on the surface corrosion of AlN substrates. The variables studied were pH, aqueous vs. organic solutions, prior surface condition, and time (up to 3.6 Ms or 42 days). The solvents tested were hydrochloric acid (HCl) with pH values ranging from 2 to 5, sulfuric acid (H₂SO₄) with pH values ranging from 2 to 5, sodium hydroxide (NaOH) with pH values ranging from 8 to 12, 1 M citric acid, oleic acid, Micro-90, methanol, ethanol, isopropanol, acetone, and deionized water. Three types of surface reaction behavior were observed in this study. The substrates either showed no reaction (HCl pH = 2, methanol, ethanol, isopropanol, acetone, citric acid, and oleic acid), slight corrosion without spalling (Micro-90, HCl pH = 3, H₂SO₄pH = 3), or they were severely corroded and spalled (HCl pH = 5, H₂SO₄pH = 5, all NaOH solutions, and deionized water).     

AlN陶瓷的薄膜金属化及其与金属的焊接研究

针对AlN陶瓷在微波管中的应用特点 ,采用磁控溅射镀膜方法对AlN陶瓷进行表面金属化 ,并与无氧铜焊接 ,测试焊接体的抗拉强度并对陶瓷 金属接合界面进行了微观分析。     

Al2 O3 陶瓷表面金属化

为了提高陶瓷与金属覆盖层的结合强度,较系统地研究了Al2O3(96%)陶瓷表面金属化过程,研究内容包括:陶瓷的表面刻蚀、表面催化、化学沉积条件等,综合分析了金属化层与陶瓷基体之间结合强度的影响因素,研究发现在熔融的NaOH浴中,可获得最佳刻蚀表面形貌,Al2O3(96%)陶瓷基片上化学镀层的最佳结合强度为25.0～32.5 MPa.为开发性应用提供了建设性的意见.     

稳健稀疏反褶积方法研究

地震勘探数字信号处理领域中一个重要而基本的方法是地震信号的反褶积方法.在地震数据的采集过程中往往会遇到异常点的干扰,这种干扰严重影响了利用反褶积方法对真实反射系数与地震子波的重构效果.本文在Canadas等人提出的针对高斯噪声的贝叶斯反褶积数学框架的基础之上,提出一种能够克服异常点干扰的稳健稀疏反褶积方法.新方法针对具有重尾分布的异常点噪声与稀疏的反射系数建模,并使用交替迭代与线性规划的算法求解.最后,通过实验证明该方法在克服异常点噪声的基础上,能实现对地震子波与反射系数的同步估计,所得到的估计有效地消除了重尾分布异常点噪声的影响,提高了地震信号反褶积处理的精度.这也能证明所提算法是收敛的,并且模型是有效的.     

玻璃纤维增强双马来酰亚胺树脂基复合材料表面金属化的研究

采用化学方法对玻璃纤维增强双马来酰亚胺树脂基复合材料的表面进行镀前处理,以提高金属化镀层的质量。金属化工艺过程包括：材料的镀前处理,化学沉积铜,电镀铜和电镀光亮镍。实验结果表明,复合材料表面的粗化工艺对于镀层的质量具有重要的作用,经过适当的化学处理,使材料表面具有均匀适当的粗糙度,同时改善表面的亲水性,提高表面对活性钯催化剂的均匀吸附能力,能够提高镀层与基体的结合力。     

Enhancing CNN for Forensics Age Estimation Using CGAN and Pseudo-Labelling

Age estimation using forensics odontology is an important process in identifying victims in criminal or mass disaster cases. Traditionally, this process is done manually by human expert. However, the speed and accuracy may vary depending on the expertise level of the human expert and other human factors such as level of fatigue and attentiveness. To improve the recognition speed and consistency, researchers have proposed automated age estimation using deep learning techniques such as Convolutional Neural Network (CNN). CNN requires many training images to obtain high percentage of recognition accuracy. Unfortunately, it is very difficult to get large number of samples of dental images for training the CNN due to the need to comply to privacy acts. A promising solution to this problem is a technique called Generative Adversarial Network (GAN). GAN is a technique that can generate synthetic images that has similar statistics as the training set. A variation of GAN called Conditional GAN (CGAN) enables the generation of the synthetic images to be controlled more precisely such that only the specified type of images will be generated. This paper proposes a CGAN for generating new dental images to increase the number of images available for training a CNN model to perform age estimation. We also propose a pseudo-labelling technique to label the generated images with proper age and gender. We used the combination of real and generated images to train Dental Age and Sex Net (DASNET), which is a CNN model for dental age estimation. Based on the experiment conducted, the accuracy, coefficient of determination (R2) and Absolute Error (AE) of DASNET have improved to 87%, 0.85 and 1.18 years respectively as opposed to 74%, 0.72 and 3.45 years when DASNET is trained using real, but smaller number of images.     

Surface Preparation of Aluminum Nitride for Metallization: Effect of Temperature on Surface Reactivity

The current aqueous cleaning step in the surface preparation of aluminum nitride (AlN) prior to metallization causes performance and reliability issues for the substrates used for microelectronic packaging due to surface reactions. These issues limit the use of AlN and its replacing of BeO, an environmentally hazardous material currently used. The aim of this investigation was to determine the effects of different solutions on the surface of AlN substrates under varying conditions at times up to 2419.2 ks (28 days). Concentration of the solutions, temperature, and immersion time were varied for the AlN samples in the solutions. Both elevated temperatures (50°C and 90°C) and low temperatures (5°C) were investigated.

Four general types of behavior were observed: minor changes in average surface roughness and microstructure, linear change in average surface roughness and pitted grains, nonlinear change in average surface roughness and product formation on AlN surface, and miscellaneous change in average surface roughness with surface product formation.

The surface roughening kinetics were very complex due to changes in both the reaction product morphology and reaction mechanism with temperature, solvent, and pH for a specific solvent. Minor changes in average surface roughness and microstructure were observed for HCl pH = 5, H₂ SO₄ pH = 5, NaOH pH = 8, NaOH pH = 10, NaOH pH = 12, deionized water and Alfred tap water at 5°C, HCl pH = 3 and oleic acid at 50°C and citric acid and oleic acid at 90°C. Linear changes in average surface roughness and pitted grains were observed for HCl pH = 2 and H₂SO₄ pH = 3 at 50°C and HCl pH = 2, H₂SO₄ pH = 3, and deionized water at 90°C. Non-linear change in average surface roughness and product formation on AlN surface was observed for HCl pH = 5, NaOH pH = 8 and Alfred tap water at 50°C and HCl pH = 5 and H₂SO₄ pH = 2 at 90°C. Miscellaneous changes in average surface roughness with surface product formation were observed for H₂SO₄ pH = 2, H₂SO₄ pH = 5, NaOH pH = 10, NaOH pH = 12, citric acid, Micro-90 and deionized water at 50°C and HCl pH = 3, H₂SO₄ pH = 5, NaOH pH = 8, NaOH pH = 10, NaOH pH = 12, Micro-90 and Alfred tap water at 90°C.