Numerical Analysis of Two-Phase Flow in a Micro-Hydraulic Turbine

Tiejun Chen,Shuaihui Sun,Ye Pang,Pengcheng Guo 한국유체기계학회 2019 International journal of fluid machinery and syste Vol.12 No.4

In many industrial processes such as natural gas processing, seawater desalination et al, there is a large amount of liquid waste pressure energy, which can be converted into electric or mechanical energy by a micro-hydraulic turbine. But its efficiency and stability will deteriorate when the working fluid contains gas. Aiming at improving the hydraulic efficiency of the micro-turbine, the three-dimensional transient two-phase flow model which adopts the shear-stress-transport (SST) model as turbulence model and Eulerian-Eulerian model as two-phase flow model with the commercial code ANSYS-CFX was established and solved to study the two-phase flow characteristics of the micro-hydraulic turbine with fixed guide vane openings under different gas volume fraction (GVF). The numerical simulation results show that the power and efficiency of the turbine decrease by 24% and 21% respectively when the inlet GVF is 0.20. The area of low pressure in each impeller passage increases with the inlet GVF, and the pressure difference between the working and suction surface of the blade decreases. The GVF distribution in impeller passage is asymmetrical, which is caused by the gas-liquid separation in the volute. At the left lower part of impeller passage, the gas accumulates at the suction surface of the blades under the Coriolis force and its GVF reaches 90%. The asymmetrical GVF distribution will result in the asymmetrical pressure distribution in the impeller passage, which leads to the imbalance force on the blades. Moreover, the asymmetrical GVF distribution will reinforce the positive incidence at the lower part of impeller passage, and weaken it at the upper right impeller passage. The volute should be redesigned to ensure the uniform gas distribution at the inlet of each impeller passage.

ResNet과 Unet을 결합한 딥러닝 모델을 이용한 분광 신호에서 ROI 검출

진철군(Tiejun Chen),손성재(SeongJae Son),박아론(Aaron Park),백성준(Sung-June Baek) 한국산학기술학회 2021 한국산학기술학회논문지 Vol.22 No.12

본 연구에서는 딥러닝 기술(deep learning technology)을 이용하여 분광 신호의 ROI(region of interest)를 찾는 방법을 제안한다. 제안한 방법은 모의실험 데이터로 학습된 딥러닝 모델을 이용하여 분광 신호의 ROI를 검출하는 방법이다. 분광 신호의 피크는 물질의 물리 화학적인 정보를 포함하고 있으므로 정확한 피크 검출은 분석 시스템의 성능에 영향을 미치는 중요한 과정이다. 지금까지 가장 많이 사용되는 방법은 진폭을 기반으로 피크 검출을 진행하는 것이다. 하지만 이런 방법들은 전처리 과정을 포함하거나 분광 신호에 따라 파라미터를 육안 검사로 선택하여 추정하므로 복잡하고 주관적이다. 이러한 문제점 개선을 위해 딥러닝 모델을 통해 분광 신호의 ROI 검출을 수행하였다. 제안한 방법은 전처리 과정이 없고 파라미터를 설정하지 않아도 되는 장점을 갖는다. 또한 검출한 ROI에 따라 분광 신호에 후처리(post-processing)를 수행하여 피크를 얻을 수 있다. 디폴트 손실 함수에 3만개 테스트 데이터를 적용하여 얻은 손실값을 통해 성능 평가를 수행하였다. 제안된 ResNet과 Unet을 결합한 딥러닝 모델은 일반적인 컨볼루션 신경망(CNN: Convolutional Neural Network), ResNet, 그리고 Unet에 비해 각각 76.5%, 69.8%, 5.9%의 성능 향상을 보였으며, 실제 라만 분광 신호의 ROI 검출에도 효과적으로 적용될 수 있음을 확인하였다. This study proposes a method to find the ROI (region of interest) of spectral signals using deep learning technology. The proposed method detects the ROI of spectral signals using a deep learning model trained with simulated data. Since the peak of the spectral signal contains physical and chemical information of the substance, accurate peak detection is an important process affecting the performance of the analyzed system. The widely used method for peak detection is the one based on the amplitude. However, this method is complex and subjective because it involves pre-processing or select and estimate parameters using visual inspection according to spectral signals. To overcome this problem, ROI detection of the spectral signal was performed through a deep learning model. The proposed method has the advantage of requiring no pre-processing and parameter setting. In addition, a peak may be obtained by performing post-processing of the spectral signal according to the detected ROI. Performance evaluation was performed through loss values obtained by applying 30,000 test data to the custom loss function. The proposed deep learning model combining ResNet and Unet showed performance improvements of 76.5%, 69.8%, and 5.9% compared to the general convolutional neural network (CNN), ResNet, and Unet, respectively. It was also confirmed that the proposed method could be effectively applied to measured spectral signals.

DDAB-MODIFIED TPGS-b-(PCL-ran-PGA) NANOPARTICLES AS ORAL ANTICANCER DRUG CARRIER FOR LUNG CANCER CHEMOTHERAPY

TIEJUN ZHAO,HEZHONG CHEN,LIXIN YANG,HAI JIN,ZHIGANG LI,LIN HAN,FANGLIN LU,ZHIYUN XU 성균관대학교(자연과학캠퍼스) 성균나노과학기술원 2013 NANO Vol.8 No.2

Oral chemotherapy is a great way to cancer treatment because it is less stressful being that the patient will have less hospital visits and can still maintain a close relationship with health care professionals. In this research, three types of nanoparticle formulation from commercial PCL and self-synthesized TPGS-b-(PCL-ran-PGA) diblock copolymer were fabricated for oral delivery of antitumor agents, including DDAB-modified PCL nanoparticles, unmodified TPGS-b-(PCL-ran-PGA) nanoparticles and DDAB-modified TPGS-b-(PCL-ran-PGA) nanoparticles. Firstly, the TPGS-b-(PCL-ran-PGA) diblock copolymer was synthesized and characterized. DDAB was adopted to increase retention time at the cell surface, thus increasing the chances of nanoparticle uptake by the gastrointestinal mucosa and improving drug absorption. The TPGS-b-(PCL-ran-PGA) nanoparticles were found by FESEM of spherical shape and around 200 nm in diameter. The surface charge of TPGS-b-(PCL-ran-PGA) nanoparticles was reversed from anionic to cationic after DDAB modification. The DDAB-modified TPGS-b-(PCL-ran-PGA) nanoparticles have significantly higher level of the cell uptake than that of DDAB-modified PCL nanoparticles and unmodified TPGS-b-(PCL-ran-PGA) nanoparticles. In vitro cell viability studies showed advantages of the DDAB-modified TPGS-b-(PCL-ran-PGA) nanoparticles over Taxotere® in terms of cytotoxicity against A549 cells. In conclusion, oral chemotherapy by DDAB-modified TPGS-b-(PCL-ran-PGA) nanoparticle formulation may provide a promising outcome for lung cancer patients.

Cabriole-legged Furniture in the Style of Song and Yuan Dynasties

Liu Tiejun,Zhang Chen 국민대학교 동양문화디자인연구소 2010 Journal of Oriental Culture&Design Vol.2 No.2

Song and Yuan dynasties is a booming period of Chinese classical furniture, and the cabriole legs became a fashion style of the furniture of that time. In this article, we use the triple proof method of images, real objects and literature to give a comprehensive analysis of the basic forms, structures and ornament features of the cabriole-Iegged furniture in the Song and Yuan dynasties, and, by taking full advantage of the rich information offered by the image data and by analyzing the relationship between furniture and human behavior, and between furniture and living space, we try our best to restore the usage mode of the furniture.

Media-aware quantitative trading based on public Web information

Li, Qing,Wang, Tiejun,Gong, Qixu,Chen, Yuanzhu,Lin, Zhangxi,Song, Sa-kwang Elsevier 2014 Decision Support Systems Vol.61 No.-

ResNet-합성곱 오토인코더 기반 신경망을 이용한 스펙트럼 데이터 압축

손성재(SeongJae Son),진철군(Tiejun Chen),박아론(Aaron Park),백성준(Sung-June Baek) 한국산학기술학회 2021 한국산학기술학회논문지 Vol.22 No.12

본 논문에서는 스펙트럼 저장 시 데이터용량을 줄이기 위해 합성곱 오토인코더(convolutional autoencoder) 구조에 ResNet(Residual Neural Network) 알고리즘을 적용한 스펙트럼 데이터 압축 신경망을 제안한다. 최근 분광법(spectroscopy)의 적용 분야가 넓어짐에 따라 스펙트럼 데이터베이스가 대용량화되어 효율적인 전송이 어렵고 많은 저장 공간을 필요로 한다. 이러한 대용량의 데이터베이스를 효율적으로 관리하기 위해 데이터 압축을 수행한다. 기존 데이터 압축에 주로 사용되는 PCA(Principal Component Analysis)는 주성분의 개수에 따라 압축률이 결정된다. 주성분 개수가 적을수록 압축률은 높아지지만 정보 손실이 보다 쉽게 발생하기 복원 시 원본 데이터와의 크게 오차가 발생한다. 이러한 한계점을 극복하기 위해 본 논문에서는 제안한 신경망인 CAER(Convolutional AutoEncoder+ResNet)을 통하여 데이터 압축을 수행하였다. 신경망 학습은 실제 스펙트럼 데이터를 묘사하여 생성한 모의실험 데이터를 통해 수행하였다. CAER 신경망의 성능 검증을 위해 라만 스펙트럼을 PCA와 신경망을 통하여 75%, 87.5%, 93.75%의 압축률로 압축과 복원을 수행한 후 각각의 결과를 비교 분석하였다. 원본과 복원 데이터의 오차 비교를 하였을 때 CAER 신경망은 PCA보다 평균 94.2%의 낮은 오차를 보인다. 이 결과를 통해 CAER 신경망이 스펙트럼 데이터 압축에 효과적으로 적용될 수 있음을 확인하였다. In this paper, we propose a spectrum compression neural network that applied the ResNet (Residual Neural Network) algorithm to the convolutional autoencoder structure to reduce data capacity requirement in storing the spectrum. Recently, as the field of application of spectroscopy widens, the spectrum database is becoming larger, making efficient transmission difficult and requiring large amount of storage. Therefore, data compression is performed to manage large amounts of data efficiently. In PCA (Principal Component Analysis), which is mainly used for data compression, the compression ratio is determined by the number of principal components. As the number of principal components decreases, the compression rate increases, but at the same time, it is easier for information loss to occur. Hence, errors occur between reconstruction and the raw spectrum. To overcome these limitations, we perform compression through the proposed CAER (Convolutional AutoEncoder+ResNet) network. The training of the network was performed through simulated data describing the real spectrum. To verify the performance of the CAER network, the Raman spectrum was compressed and reconstructed at compression rates of 75%, 87.5%, and 93.75% through the PCA and CAER networks. Comparing the errors between raw and reconstructed data, the CAER network shows an average error of 94.2% lower than that of the PCA. The results obtained confirm that the CAER network can be effectively applied to spectrum compression.

EFFICIENT PRODUCTION OF BIO-GASOLINE FROM SUGAR POLYOLS OVER RU-MOLYBDENUM OXIDE CATALYSTS

Songbai Qiu,Yujing Weng,Lungang Chen,Longlong Ma,Qi Zhang,Tiejun Wang 한국신재생에너지학회 2015 AFORE Vol.2015 No.11

Spontaneous wrinkling pattern of a constrained thin film membrane

Yan, Yuan,Wang, Binglei,Yin, Jie,Wang, Tiejun,Chen, Xi Springer-Verlag 2012 APPLIED PHYSICS A MATERIALS SCIENCE AND PROCESSING Vol.107 No.4

Dietary Supplementation with Acanthopanax senticosus Extract Modulates Gut Microflora in Weaned Piglets

Yin, Fugui,Yin, Yulong,Kong, Xiangfeng,Liu, Yulan,He, Qinghua,Li, Tiejun,Huang, Ruilin,Hou, Yongqing,Shu, Xugang,Tan, Liangxi,Chen, Lixiang,Gong, Jianhua,Kim, Sung Woo,Wu, Guoyao Asian Australasian Association of Animal Productio 2008 Animal Bioscience Vol.21 No.9

This study was conducted to investigate the effects of Acanthopanax senticosus extract (ASE) as a dietary additive on gut microflora in weaned piglets. A total of sixty pigs were weaned at 21 d of age (BW = $5.64{\pm}0.23kg$) and allocated on the basis of BW and litter to three dietary treatments in a randomized complete block design. The dietary treatments were: control group (basal diet), antibiotics group (basal diet+0.02% colistin), and ASE group (basal diet+0.1% ASE). On d 7, 14 and 28 after consuming the experimental diets, five piglets per group were sacrificed and then the contents from the jejunum, ileum and cecum were collected to determine changes in the microbial community by using a polymerase chain reaction-denaturing gradient gel electrophoresis (PCR-DGGE) technique and estimating the contents of Lactobacillus and E. coli by in vitro culturing methods. The results showed that the ASE promoted the microflora diversity in the cecum. Enumeration of bacteria in the gut contents showed that the number of Lactobacillus increased (p<0.05), while that of E. coli decreased (p<0.05) when compared with the other 2 groups as the days of age progressed post-weaning. These findings suggested that the ASE, as a substitute for dietary antimicrobial products, could improve the development of the normal gut microflora and suppress bacterial pathogens, and effectively promote a healthy intestinal environment.

Rumex hanus by. Extract Protects Against Chronic Alcohol-Induced Liver Injury in Mice

Meizi Piao,Fengwu Wang,Lingyue Shan,Yang Deng,Tiejun Chen 한국식품영양과학회 2022 Journal of medicinal food Vol.25 No.7

Alcoholic liver disease (ALD) has become a global health problem. The hepatoprotective effects of bioactive ingredients extracted from Rumex hanus by. on chronic alcoholic liver injury was investigated for the first time. The extract from R. hanus by. (ERHB) was obtained by 70% ethanol extraction, and the endotoxin antagonism rate of ERHB was 88.94 ± 1.24% in vitro. The animal experiments demonstrated that ERHB promoted hepatic function by significantly enhancing the activities of alcohol dehydrogenase and acetaldehyde dehydrogenase, and by reducing the activities of cytochrome P450 proteins, alkaline phosphatase, aspartate aminotransferase, and alanine aminotransferase. Furthermore, ERHB improved alcohol-induced dyslipidemia by regulating lipid metabolism. In addition, ERHB ameliorated the alcohol-induced liver injury by inhibiting endotoxin-caused inflammation. Seven compounds with antagonistic activity on endotoxin were identified in ERHB. These results demonstrated that ERHB had protective effects on ALD and if the results can be confirmed in humans, it might be useful as a functional food supplement for ALD treatment.