공력저항에 따른 wide area munition의 동역학적 거동 해석

홍준호 고려대학교 대학원 2017 국내석사

ABSTRACT Analysis of Dynamic Behavior of a Wide Area Munition with Aerodynamic Effect Name: Junho Hong Department: Mechanical Engineering Thesis Advisor: Woo Chun Choi The wide area munition (WAM) investigated in this study is munition that automatically attacks a target. The WAM discharged from platform flies with precession. An IR sensor attached to the WAM draws a specific trajectory on the ground. When a target is detected, the WAM attacks the target from the air. Therefore, the trajectory of IR sensor is important for accurate attack. Aerodynamic effect and precession are important for the trajectory of IR sensor. In this study governing equations about precession were derived, and the WAM pose was determined using numerical analysis method. Also CFD analysis was done to find the drags and moments for each WAM pose for different rotation speeds and wind speeds. Finally, the trajectory of IR sensor and flight are determined with aerodynamic effect considered. Also in order to find tendency of trajectory of IR sensor and flight, numerical analysis is conducted for many variables. It was found that the aerodynamic resistant forces affect the deviation distance and maximum distance, while the aerodynamic resistant moments affect the pose of the WAM.

Three-Component Synthesis of Quinolines Based on Radical Cascade Visible-Light Photoredox Catalysis

Jun-Ho Choi Ulsan National Institute of Science and Technology 2021 국내박사

Recently, irradiation of visible light photoredox catalyst has been widely used to drive transformation of organic molecules though the highly desirable processes. Absorption of visible light induces the long-lived excited-triplet state of the photocatalyst. The returning from this excited state to the bench state of the catalyst leads single-electron-transfer (SET), transforming the reacting substrates into radical intermediates. The radical cation/anion intermediates show their reactivity which is basically differentiated from those of electronically ground state of molecules. Multicomponent reaction (MCR) serves as a powerful tool, employing three or more simple building blocks to produce complex molecular frameworks in a single step. Given its advantages of rapid access to structural complexity, MCR has been adopted as a valuable means for the discovery of bioactive compounds. As such, photoredox catalysis has been explored in the context of MCR. However, the limited examples are mostly based on radical-polar crossover mechanisms, in which the process is in operation for initial coupling of two reactants, while a third component is incorporated via a polar process. Although this approach is useful, quenching of radical processes by redox catalysts entailing the involvement of polar processes poses limitations in the scope of coupling partners. To the best of our knowledge, MCR with three consecutive bond formation based on radical processes via visible-light photoredox catalysis has not been reported. Due to their utilities, significant efforts have been made to develop efficient synthetic methods for quinolines. However, the conventional methods rely on condensation under harsh conditions and more recent developments are limited to transitional metal-catalyzed and iodine-mediated synthesis. Here, a successful development of a new tandem radical cyclization based on visible-light photoredox catalysis enables the efficient formation of quinolines based on consecutive radical processes.

Analysis of drag coefficient based on velocity and distance ratio in the Ahmed model platoon driving

Choi Jun-Ho 한국기술교육대학교 2025 국내석사

아흐메드(Ahmed) 모델은 일반적인 자동차 형상을 상당히 단순화시킨 모델이며, 항력 메커니즘 분석을 위한 대표적인 기준 모델로 널리 사용된다. 군집주행은 차량 간 유동 상호작용을 통해 전체 항력 감소 효과를 유도할 수 있는 기술이다. 본 연구에서는 아흐메드 모델을 기반으로 속도 및 차간거리 변화에 따른 2대, 3대, 그리고 4대의 군집주행 조건에 대한 수치해석을 수행했고, RANS 기반 SST k-ω 난류모델을 적용하여 항력계수 변화를 예측했다. 이전 연구에서 보고된 실험값과의 비교를 통해 0.6%에서 5.0% 이내의 오차를 가진 타당성 검증을 통해 높은 신뢰성을 확보했다. CFD 해석 결과 차간거리가 가까울수록 항력 감소 효과가 크게 나타나고, 군집 내 차량 수가 증가할수록 연료 소비 감소율이 향상됨을 확인했다. 또한, 모든 속도 조건에서 차간거리가 멀어질수록 차량 간 유동 상호작용이 약화되며 항력계수는 유사해지는 경향을 보였고, 이를 바탕으로 속도 및 차간거리에 따른 항력계수 상관식이 제시되었다. The Ahmed model, a simplified representation of a typical vehicle shape, is widely used as a benchmark model for analyzing aerodynamic drag mechanisms. Platooning is a technique that reduces overall aerodynamic drag through flow interactions between vehicles. In this study, numerical simulations were conducted under platooning conditions involving 2, 3, and 4 vehicles, based on the Ahmed model, considering variations in velocity and inter-vehicle distance ratio. The RANS-based SST k-ω turbulence model was applied to predict changes in drag coefficients. Validation against previously reported experimental data demonstrated high reliability, with prediction errors ranging from 0.6% to 5.0%. CFD results indicated that shorter inter-vehicle distances significantly enhance drag reduction, and increasing the number of vehicles in the platoon further improves fuel savings. Across all velocity conditions, increased inter-vehicle spacing weakened flow interactions, resulting in drag coefficients that gradually converged. Based on these findings, an empirical correlation was developed to express the drag coefficient as a function of velocity and inter-vehicle distance ratio.

Ct-1, one of the Arabidopsis thaliana ecotypes, has higher salinity tolerance than Col-0

Jun Ho Choi 고려대학교 대학원 2023 국내석사

농업 생산량의 약 40%는 충분한 양의 강수가 공급되지 못해서 관개에 의존한다. 그러나 관개용수에는 미네랄이 소량 함유되어 있을 뿐만 아니라 지속적인 관개로 인해 토양 내 염분 함량이 높아져 결국 농작물의 생산량이 감소하게 된다. 따라서 식물의 염분 저항성 메커니즘을 이해하기 위한 많은 연구가 수행되어 왔다. 우리는 애기장대의 여러 생태형 중 하나인 Ct-1이 Col-0보다 높은 염분에 내성이 있다는 것을 확인했다. Ct-1은 Col-0에 비해 활성산소에 의한 손상이 적었으나 안토시아닌과 같은 항산화물질은 축적되지 않았다. 본 연구에서는 Ct-1이 높은 염도에서 Col-0에 비해 싹에 더 많은 K+ 이온을 축적할 수 있다는 것을 발견했다. K+ 이온을 확보하는 것은 식물의 내염성에 매우 중요하다. 또한 Ct-1의 NO3- 함량은 K+ 이외에도 NO3 ¬ 함량이 Col-0에 비해 높게 나타났는데 이는 K+ 이온의 증가와 관련이 있는 것으로 판단된다. NRT1.5와 NPF2.3과 같은 질산염 관련 유전자가 높은 염분에 반응하여 Col-0보다 Ct-1에서 더 많이 발현되는 것을 확인하였다. 결과적으로 질산염과 칼륨 사이의 수송 능력의 차이가 식물의 염 스트레스에 대한 저항성을 향상시킬 수 있다는 것을 의미한다. Approximately 40% of the total agricultural output is dependent on irrigation, as not all agricultural areas receive adequate amounts of rainfall. However, irrigation water not only contains a small amount of minerals, but also increases salt content in the soil due to continuous irrigation, which eventually leads to a decrease in crop production. Therefore, many studies have been conducted to understand the mechanisms of salinity resistance in plants. We have confirmed that Ct-1, one of several ecotypes of Arabidopsis, is more resistant to high salinity than Col-0. Ct-1 was less damaged by reactive oxygen species (ROS) than Col-0 but did not accumulate antioxidants such as anthocyanins. In this study, we found that Ct-1 can accumulate more K+ ions in shoots compared to that by Col-0 under high salinity. Securing K+ ions is very important for the salt resistance in plants. In addition to K+, the NO3– content in Ct-1 was also higher than that of Col-0, which seems to be related to the increase of K+ ions. It was confirmed that nitrate-related genes such as NRT1.5 and NPF 2.3 were more expressed in Ct-1 than in Col-0 in response to high salinity. Our results imply that the difference in transport capacity between nitrate and potassium can improve resistance to salt stress in plants.

(A) study on high-power transmission based on electrical breakdown analysis for several-hundred-kilovolt and gigawatt nano-pulse in microwave coherent source array

최준호 Gwangju Institute of Science and Technology 2015 국내박사

Alleviation of high cholesterol-induced hepatic steatosis by lactobacillus acidophilus ATCC 43121

최준호 Graduate School, Korea University 2014 국내석사

Nonalcoholic fatty liver disease (NAFLD) is a common chronic liver disease that ranges in severity from simple steatosis to cirrhosis. Now NAFLD is considered to be associated with hepatic metabolic disorders, resulting in overaccumulation of fatty acids/triglycerides and cholesterol. Especially, cholesterol has an important role of development of NAFLD. In some studies of nutritional animal models, the accumulation of cholesterol rather than fatty acids/triglycerides plays a critical role in this progression. It has been reported that probiotics has the ability to lower plasma cholesterol levels when animals are fed with it. Many people, heretofore, focused on how to decrease cholesterol levels in serum and overlook the role of cholesterol for hepatosteatosis. The purpose of the present study was to evaluate the alleviation of hepatic steatosis induced by high cholesterol diet with probiotics. Lactobacillus acidophilus ATCC 43121, which has high ability of cholesterol assimilation, was used to improve the liver steatosis. This strain has excellent acid tolerant character and showed high adhesion activity to HT-29 cells compared to GG strain. This indicated that 43121 strains has potential to positive effect in human intestine when intake it. And we checked whether our strain can modulate the cholesterol metabolism using HepG2 cells, cultured human liver cell, but it has no effect to HepG2 cells. However, because our strain has high ability of cholesterol assimilation, potential effect of L. acidophilus ATCC 43121 on plasma cholesterol level and liver steatosis was evaluated in C57BL/6N mice. Six week old male mice were fed a commercial diet (cholesterol-free diet, negative control, NC), cholesterol-enriched diet 28 days. We investigated the effects of L. acidophilus ATCC 43121 on cholesterol levels in serum and liver steatosis. And also we checked hepatic mRNA expression related lipid metabolism and TG contents in liver. The results showed that our strain could decrease levels of LDL cholesterol levels (p<0.05) and total triglyceride (TG) in liver (p<0.05), but the levels of total cholesterol and HDL-C in plasma was unchanged (p<0.05). In addition, the result of histologic analysis showed that high cholesterol group supplemented with probiotic had improved liver compared with high cholesterol groups. This study suggested that probiotics improve hepatic steatosis by reducing LDL-c and TG contents in liver. In conclusion, these results suggest that Lactobacillus acidophilus ATCC 43121 have the potential to be explored as a probiotic in the management of non-alcoholic fatty liver disease.

Development of new purification and gene transfer method to intestinal parasites for immune disorder therapeutic agents

최준호 부산대학교 대학원 2019 국내박사

최근, T. suis (whipworm) 충란을 사용하여 자가 면역 질환을 치료하려는 시도가 있어 왔다. 돼지의 대변에서 대량으로 충란을 얻는 방법이 T. suis 암컷 자궁에서 충란을 추출하는 것보다 효율적이다. 그러나, 기존의 sieve 방법을 사용하여 많은 양의 대변을 처리하는 것은 어렵다. 본 연구에서는 T. suis 대변에서 충란을 효율적으로 수집하는 방법을 제안한다. Mesh를 사용하여 100 g의 대변에서 충란을 모았을 때, 25 μm에서 65.7% (± 5.7; mean ± SD)의 충란을 수집했다. 에틸 아세테이트와 인공 위액을 사용하는 SGF 방법에서는 sucrose 수용액에서 91.4% (± 1.3; mean ± SD)의 충란을 수집했다. SGF 방법을 이용하면 충란의 수집은 기존의 sieve 방법을 사용한 방법보다 60분 더 빨랐다. SGF 실험에 사용된 인공 위액은 pH 1-2의 강산성 시약이다. 따라서, 충란의 larva는 급속한 pH 변화를 주었으며, 충란의 embryonation이 유도되었다. 그 결과, 인공 위액 자극 후 2개월 만에 39.5% (± 27.4; mean ± SD)의 충란이 embryonation을 완료하여 감염형의 larva가 충란 내에 형성되었다. 이 연구의 결과로 개발된 SGF 방법을 사용하면 신속하고 효율적으로 대변에서 충란을 수집할 수 있다. 또한, 충란의 정제 후 임의의 유해한 시약 처리없이 충란 내에 larva를 유도할 수 있다. SGF 방법은 T. suis eggs를 이용한 향후의 연구에 도움이 될 것으로 기대된다. 녹아웃 마우스 생산 및 C. elegans의 변형에 널리 사용되는 microinjection 기술은 유전자 조작에 유용한 방법이다. 그러나, T. spiralis에 대한 microinjection 기술은 아직 확립되지 않았다. T. spiralis는 인체에 감염될 수 있는 중요한 기생 선충이다. 따라서 이 기생 선충에 대한 유전 연구는 필수불가결하다. 실험에 사용된 T. spiralis L1 larva는 4주 전에 T. spiralis L1 larva에 감염된 C57BL/6 마우스의 근육을 소화시켜 얻은 것이다. Microinjection에 사용된 플라스미드는 항 종양 괴사 인자-알파 (anti TNF-α) 유전자 서열을 pPV230.13에 삽입하여 제작된 치료용 벡터이다. Glass Capillary, GD-1 및 PC-10 Puller를 사용하여 microinjection에 사용할 인젝터 피펫을 제작하였다. 제작된 플라스미드는 IM-300 Microinjector, NARISHIGE를 사용하여 T. spiralis L1 larva의 genital primordium에 주입되었다. pPV230.13_anti TNF-α가 주입된 T. spiralis L1 larva의 100 마리 중에 90 마리에서 형광이 관찰되었다. 또한 주입된 플라스미드가 T. spiralis L1 larva의 체내에서 형광을 3일 동안 유지하는 것도 확인했다. 추가적으로, 장관 내 기생충인 T. suis의 성충과 충란에 microinjection을 실시하였다. 그 중에서도 성충에서만 형광의 발현이 확인되었다. 또 다른 문제는 T. suis의 adult worm은 시험관 내에서 장기간 생존할 수 없다는 점, 성충의 상태로는 돼지에 재 감염할 수 없는 문제가 있었다. 이러한 점은 유전적으로 변형된 T. suis를 확보하기 어렵기 때문에 다른 접근법이 필요하다. 예를 들어, T. suis 성충과 충란에 미세 펄스를 주는 방법이 있다. 이 방법은 T. suis 성충 및 충란을 손상시키지 않고 체내로 플라스미드를 전달하는 방법이다. 생체 내 면역 반응은 염증성 사이토카인 TNF-α 및 IL-17A의 농도가 anti-TNF-α 유전자가 삽입된 플라스미드를 접종한 T. spiralis L1 larva에 감염된 군에서 낮아지거나 유사한 수준을 보였다. 이는 마우스 비장과 MLN에서의 TNF-α 농도가 anti-TNF-α 유전자가 삽입된 플라스미드 접종 실험군에서 양성 대조군인 T. spiralis 감염 군보다 낮은 수준으로 사이토카인이 발현된 결과이다. 이것은 T. spiralis에 삽입된 anti-TNF-α 유전자가 발현되어 생성된 단백질이 마우스 TNF-α에 대해 항 TNF-α로서 작용한다고 생각된다. 이러한 현상은 마우스 장 조직 내의 mRNA 수준에서도 동등하게 확인된다. IL-4와 IL-17A는 양성 대조군인 T. spiralis과 anti-TNF-α 유전자가 삽입된 플라스미드 접종 실험군의 마우스 생체 내에서 mRNA 발현 수준과 사이토카인 발현 수준은 서로 다르게 나타났다. 이것은 생쥐에 감염된 T. spiralis 100ea의 수가 마우스에서 면역 반응을 유도하기에는 불충분하다고 생각된다. 또 다른 가능성은 소장에서 T. spiralis 성충의 수가 감염 2주 후에 점차적으로 체외로 방출되고나서 마우스의 소장, 비장 그리고 MLN의 세포 수준에서 발현되는 mRNA의 양과 분비되는 사이토카인의 양이 감염이 지속적으로 유지되는 4주 후와는 차이가 난다는 점이다. 그러나, ani-TNF-α 유전자가 삽입된 T. spiralis가 감염된 마우스에서 TNF-α의 수준은 mRNA 및 사이토카인 모두 낮고, 이 결과는 T. spiralis에 삽입된 ani-TNF-α 플라스미드가 제대로 작동한다는 것을 보여준다. 이번 연구에서 사용된 플라스미드의 설계, microinjection을 수행하는 방법들은 차후에 T. spiralis의 유전적 변형 연구에 사용될 수 있는 유용한 도구가 될 것으로 생각된다. Recently, attempts have been made to treat autoimmune diseases using Trichuris suis (whipworm) eggs. It is more efficient a method that can obtain in large quantities by collecting from the feces of the porcine hosts rather than extracting the eggs from the female uterus. However, it is difficult to handle large amounts of feces using current methods. In the present study, we propose a method to collect the eggs from the bulk of the feces efficiently. Collecting the eggs by washing meshes with 100 g of feces yielded 65.7% (±5.7) of the eggs at 25 µm (mean ± SD). Our method, which uses ethyl acetate and simulated gastric fluid, collected 91.4% (±1.3) of the eggs into the separated aqueous solution. Egg collection using our method was 60 mins faster than that using the conventional method. The simulated gastric fluid used in the experiment is a strongly acidic reagent with a pH of 1–2. Thus, embryonation of the eggs induced when they experienced a rapid pH change. As a result, 39.5% (±27.4) of the eggs had embryonated two months after simulated gastric fluid stimulation. Using the developed method, we could process the feces quickly and efficiently. Also, after purification, egg embryonation could be induced without any harm reagent treatment. This method is expected to be helpful for further research using Trichuris suis eggs. Microinjection technology, which is widely used for knockout mouse production and transformation of Caenorhabditis elegans (C. elegans), is advantageous for manipulating genes. However, microinjection technology for nematode parasites has not been established. Most parasites infect a specific site in the host; thus, parasites may be useful as organ-specific drug delivery systems. We choose the nematode parasite Trichinella spiralis, which is capable of infecting the intestines of the human body. In this study, we aimed to establish a microinjection technique against the parasite, express the transferred gene in the parasite, and achieve an amelioration effect by transferred gene expression. We used pPV230.13 (pPV230) as the plasmid vector for microinjection, and part of the anti-tumor necrosis factor (TNF)-α antibody coding gene with the GFP gene was inserted into the vector to create pPV230-anti-TNF-α. This vector was injected into the genital primordium of T. spiralis muscle larva using a microinjector. We observed fluorescence (90%) in the genital primordium of the larva and confirmed the presence of the gene by PCR. The expression level of TNF-α in the mouse intestine infected with T. spiralis containing pPV230-anti-TNF-α was significantly decreased compared to that in normal T. spiralis-infected mouse intestines. In conclusion, we used a microinjection technique for gene or drug transport into a nematode parasite, suggesting that drug delivery systems can be developed using parasites. Inflammatory cytokines TNF-α and IL-17A showed lower or similar levels in T. spiralis L1 larva. Also, TNF-α showed a lower level of cytokine expression than the microinjected T. spiralis infection group in the plasmid inoculation group in which the anti-TNF-α gene was inserted. This fact suggests that the anti-TNF-α gene of T. spiralis acts as an anti-TNF-α for mouse TNF-α. This phenomenon is equally confirmed at the mRNA level. The level of T. spiralis mRNA inoculated with the pPV230.13 vector was not significantly lower, but the level of cytokine was lower than that of infected T. spiralis without microinjection and microinjection affected the activity of T. spiralis. IL-4 and IL-17A showed different patterns in infected T. spiralis by insertion of pPV230.13 vector compared to infected T. spiralis without microinjection. mRNA expression levels and cytokine expression levels are different. This fact suggests that the number of mice infected T. spiralis 100ea is insufficient to induce an immune response in mice. Another possibility is that the number of T. spiralis adults in the small intestine gradually diffuses into the body after two weeks of infection and the amount of mRNA expressed at the cellular level differs from that secreted by cytokines. However, the level of TNF-α in T. spiralis with ani-TNF-α gene insertion is low in both mRNA and cytokines, and the ani-TNF-α plasmid inserted into T. spiralis works well. In this study, the methods of plasmid design and microinjection could be useful tools for next genetic modification of T. spiralis.

(The) estimation of N2O emission from environmental fundamental facility

최준호 Graduate School, Yonsei University 2010 국내석사

Study on transparent electromagnetic interference shielding using improved graphene by electroplating

Choi, Hojun Sungkyunkwan university 2019 국내석사

Recently, much attention has been paid to a flexible device due to development of the wearable device and foldable phone. However, it is reported that electromagnetic waves occur as driving pulse is applied to a touch screen panel of the flexible device and therefore the electromagnetic waves interference occurs between electronic components. High electronic conductivity, transmittance, flexibility, thin thickness are needed in the flexible device for a transparent and flexible electromagnetic wave shielding. Indium tin oxide(ITO) is used for transparent electromagnetic wave shielding due to high conductivity and transmittance but it is easily broken due to low failure strain. In addition, alternative is needed due to price increase and resource depletion of the indium. Graphene is a candidate instead of the ITO but its theoretical properties were low due to defects. Therefore, in this study, defects of the graphene were improved by using selective electro-deposition for fabrication of the transparent and flexible electromagnetic wave shielding materials. In this study, the graphene was synthesized by chemical vapor deposition and transferred on PET by thermal release tape. Ag was selected due to oxidation stability and high electronic conductivity. Parameters for optimized plating were current density (ASD), plating time and additive(PEI). A morphology of Ag particles were examined through SEM, OM and grain size and crystalline were confirmed by XRD. Ag binding energy was measured by XPS. In addition, selective plating at line defect, pin hole were confirmed through Raman spectroscope. In order to compare Ag plated graphene and defective graphene, four point probe, UV-vis, Attenuation were measured. Shielding effectiveness was measured by vector network analyzer in 1 – 18GHz.

Robust deep neural networks against limited resources, data unavailability, and adversarial attacks

최준호 Graduate School, Yonsei University 2021 국내박사

최근 심층 신경망 기법이 다양한 응용 분야에 접목됨에 따라, 심층 신경망 기법을 개발하는 환경과는 전혀 다른 특성을 지니는, 응용 분야가 적용되는 실제 환경의 특성을 이해하는 것이 중요해지고 있다. 특히, 심층 신경망을 연구하는 환경은 주로 풍부한 계산 자원과 오류가 발생하지 않는 데이터 획득 상황을 가정한다. 그러나, 실제 환경은 고비용의 계산 자원과 기술적인 오류로 인한 데이터 손실이 필연적으로 발생할 수 있으므로, 통제된 환경을 보장할 수 없다. 심층 신경망 기법을 응용 분야에 바로 접목하기 어렵게 하는 요인으로는 크게 세가지가 있다. 첫째, 심층 신경망 기법이 배포된 환경의 경우 효율성을 확보하기 위해 데이터의 정보량을 감소하는 기법을 사용할 가능성이 높으며, 이는 기법의 인식 및 재건 성능을 직접적으로 하락시킬 수 있다. 둘째, 실제 환경에서는 기술적인 오류로 인해 일부 데이터가 손실될 수 있으며, 이는 사용하는 모든 데이터 정보가 항상 존재한다고 보장할 수 없게 한다. 셋째, 심층 신경망 기법을 교란시키기 위해 악의적인 공격이 가해질 수 있으며, 이는 치명적인 보안 문제를 야기한다. 본 학위논문에서는 실제 응용 분야에서 활용될 수 있는 심층 신경망 기법들의 강인성을 위에서 언급한 요인들을 중심으로 분석하고, 이에 강인한 심층 신경망 기법을 제안한다. 연구 대상 및 방법은 다음과 같다. 우선, 행동 인식 작업에 주로 수반되는 전통적인 특징 추출 기반 기법과 심층 신경망 기반 기법에 대해 삼차원 영상 확장성(video scalability)을 바탕으로 강인성을 비교한다. 이를 통해, 심층 신경망 기반 기법이 정보량 감소에 보다 강인한 성능을 보이며, 성능을 유지할 수 있는 최적의 영상 확장성 요소 조합을 제안한다. 다음으로, 다중 모달리티(multimodal) 정보를 활용하는 인식 작업에서 수반되는 데이터 손실에 강인한 심층 신경망 기법을 제안하며, 기존의 최신 모달리티 정보 융합 방법과의 성능 비교를 통해 높은 성능을 확인한다. 마지막으로, 최근 주로 활용되는 이미지 재건 기법 중 하나인 초해상도 작업용 심층 신경망 모델의 적대적 공격(adversarial attack)에 대한 강인성을 분석한다. 이를 통해 최신 심층 신경망 기법들이 적대적 공격에 매우 취약함을 확인하며, 이를 해결할 수 있는 방어 기법을 제안한다. As deep neural networks have been widely used in various applications nowadays, it is important to understand the characteristics of the real-world environments that are quite different from the conditions where the deep network models are built. In particular, studies for deep neural networks usually assume well-controlled environments with abundant computing resources and errorless data acquisition processes. When it comes to the wild, these conditions cannot be ensured due to limited and expensive computing resources and possible technical data corruption. The main obstacles of the direct application of deep neural networks to real-world situations are as follows. First, the deployed environments may employ some processes to reduce the amount of data for efficiency, directly affecting the recognition and reconstruction performance. Second, some data may be lost due to some technical difficulties; i.e., it is hard to guarantee that data of all modalities always exist. Third, there can be some malicious attacks to fool the deep neural networks, which may cause critical security concerns. This thesis examines the robustness of deep neural networks in real-world applications regarding the obstacles mentioned above in three-fold. First, both traditional feature extraction-based and deep learning-based activity recognition methods are investigated with respect to three-dimensional video scalability dimensions. The results show that the deep learning-based approach is more robust against data loss than the traditional approaches. A thorough analysis of various scalability combinations is also shown, suggesting optimal combinations to keep performance with the reduced data rate. Second, a novel multimodal fusion approach for deep recognition methods is proposed, which can naturally handle data loss. The performance is compared with various state-of-the-art fusion approaches. The results show that the proposed approach achieves high recognition performance even when some parts of data are missing. Third, deep models for super-resolution, which is one of the widely used reconstruction tasks, are evaluated against adversarial attacks. The results show the high vulnerability of state-of-the-art methods. In addition, a defense method is proposed, which is formulated as an entropy regularization loss for model training.