• 정보과학회 논문지
• /
• 제44권7호
• /
• pp.668-673
• /
• 2017

최근 몇 년간 딥러닝(deep learning)은 음성 인식, 영상 인식, 물체 검출을 비롯한 다양한 패턴인식 분야에서 혁신적인 성능 발전을 거듭해왔다. 그에 비해 네트워크가 어떻게 작동하는지에 대한 깊은 이해는 잘 이루어지지 않고 있다. 본 논문은 효과적인 신경망 네트워크를 구성하기 위해 네트워크 파라미터들이 신경망 내부에서 어떻게 작동하고, 어떤 역할을 하고 있는지 분석하였다. Faster R-CNN 네트워크를 기반으로 하여 신경망의 과적합(overfitting)을 막는 드랍아웃(dropout) 확률과 앵커 박스 크기, 그리고 활성 함수를 변화시켜 학습한 후 그 결과를 분석하였다. 또한 드랍아웃과 배치 정규화(batch normalization) 방식을 비교해보았다. 드랍아웃 확률은 0.3일 때 가장 좋은 성능을 보였으며 앵커 박스의 크기는 최종 물체 검출 성능과 큰 관련이 없다는 것을 알 수 있었다. 드랍아웃과 배치 정규화 방식은 서로를 완전히 대체할 수는 없는 것을 확인할 수 있었다. 활성화 함수는 음수 도메인의 기울기가 0.02인 leaky ReLU가 비교적 좋은 성능을 보였다.

• Environmental Engineering Research
• /
• 제17권3호
• /
• pp.139-144
• /
• 2012

In this study, the effects of starvation on floc characteristics when treating saline wastewater using a sequencing batch reactor (SBR) were investigated. The effectiveness over 5 days of starvation for aerobic and non-aerobic strategies for maintaining the physical characteristics of floc-forming sludge and the recovery period needed to regain the initial pollutant removal efficiency were investigated. Experiment results revealed that the sludge volume index (SVI) increased and the floc size and fractal dimension decreased after starvation under both aerobic and non-aerobic conditions. Sludge settleability deteriorated faster under aerobic conditions compared to non-aerobic conditions. Under non-aerobic conditions, the SBR required less time to return to its initial pollutant removal efficiency and settleability. Floc size, fractal dimension, and SVI were observed to be fairly correlated with each other. The results demonstrated that it was better to maintain the sludge under non-aerobic rather than aerobic starvation, because it adapted to, resisted starvation and had a quicker re-start afterward.

• 한국물환경학회지
• /
• 제25권4호
• /
• pp.515-521
• /
• 2009

In this study, the applicability of $CaCO_3$ as a seed material for crystallization reaction was tested. $CaCO_3$ was ground to lesser than 425 mesh and was made to media mixed with binder. Batch experiment was to investigate the ${PO_4}^{3-}-P$ removal efficiency of different parameters such as $CaCO_3$ dosage and binder ratio, size, type and mass of media. In addition, the effect of phosphorus removal from wastewater was tested using a lab-scaled crystallization reactor. At the results of the batch test, phosphorus removals were improved with increasing $CaCO_3$ dosage and media mass but were decreased with increasing media size. Moreover, phosphorus removals were influenced by specific surface area but media type. The average T-P and ${PO_4}^{3-}-P$ removal efficiency in a lab-scaled crystallization reactor with $CaCO_3$ media for wastewater were shown to be 60.2% and 60.3% for 18 days of operation time.

• 산업경영시스템학회지
• /
• 제43권2호
• /
• pp.120-125
• /
• 2020

Supply chain managers seek to achieve global optimization by solving problems in the supply chain's business process. However, companies in the supply chain hide the adverse information and inform only the beneficial information, so the information is distorted and cannot be the information that describes the entire supply chain. In this case, supply chain managers can directly collect and analyze supply chain activity data to find and manage the companies described by the data. Therefore, this study proposes a method to collect the order-inventory information from each company in the supply chain and detect the companies whose data characteristics are explained through deep learning. The supply chain consists of Manufacturer, Distributor, Wholesaler, Retailer, and training and testing data uses 600 weeks of time series inventory information. The purpose of the experiment is to improve the detection accuracy by adjusting the parameter values of the deep learning network, and the parameters for comparison are set by learning rate (lr = 0.001, 0.01, 0.1) and batch size (bs = 1, 5). Experimental results show that the detection accuracy is improved by adjusting the values of the parameters, but the values of the parameters depend on data and model characteristics.

• 한국생물공학회:학술대회논문집
• /
• 한국생물공학회 2003년도 생물공학의 동향(XII)
• /
• pp.378-381
• /
• 2003

질소, 인 등의 영양소 제거를 위한 고정화 미생물을 이용한 새로운 BNR 시스템 개발을 위해 탈질능력이 뛰어난 균주를 순수분리하고 여러 담체를 사용하여 제조된 활성 bio-bead의 특성을 알아보았다. 4종류의 탈질 능력이 뛰어난 균주를 순수분리 하였고, 12% PVA 담체로 만든 bead가 가장 좋은 탈질능력을 보였다. Cell loading 농도는 200 mg/ml일 때 bead의 활성이 최적을 나타내었으며, bead의 size에 의한 탈질효율에는 유의적 차이가 없었다. 제조된 bead는 약 4 batch실험 후 최대 탈질율을 나타내었으며, 약 25번의 batch실험까지는 탈질능력을 그대로 유지하고 cell leaking 양도 약 $10^3$ CFU/ml 정도 였으나, 그 이후로는 bead 형태도 변하면서 cell leaking이 증가하고 탈질능력도 감소되었다.

• Management Science and Financial Engineering
• /
• 제13권2호
• /
• pp.47-77
• /
• 2007

This paper considers a scheduling problem where a customer orders multiple products(jobs) from a production facility. The objective is to minimize the sum of the order(batch) completion times. While a machine can process only one job at a time, multiple machines can simultaneously process jobs in a batch. Although each job has a unique processing time, we consider the case where batch processing times are identical. This simplification allows us to develop heuristics with improved performance bounds. This problem was motivated by a real world problem encountered by foreign electronics manufacturers. We first establish the complexity of the problem. For the two parallel machine case, we introduce two simple but intuitive heuristics, and find their worst case relative error bounds. One bound is tight and the other bound goes to 1 as the number of orders goes to infinity. However, neither heuristic is superior for all instances. We extend one of the heuristics to an arbitrary number of parallel machines. For a fixed number of parallel machines, we find a worst case bound which goes to 1 as the number of orders goes to infinity. Then, a tighter bound is found for the three parallel machine case. Finally, the heuristics are empirically evaluated.

• 한국생물공학회:학술대회논문집
• /
• 한국생물공학회 2000년도 추계학술발표대회 및 bio-venture fair
• /
• pp.287-290
• /
• 2000

회분배양에서 교반속도와 통기량에 대한 영향을 살펴본 결과 400 rpm과 1.0 vvm의 조건에서 균체량은 25.1 g/L, 다당체는 2.5 g/L로 가장 높게 생성되었다. 회분배양을 기초로 균체량과 다당체 생성을 높이기 위해 glucose를 대수기에 공급한 결과 배양 6일째 균체량은 29,2 g/L, 다당체 생성은 3.3 g/L로 회분배양보다 우수한 결과를 보였다.

• 한국축산시설환경학회지
• /
• 제2권2호
• /
• pp.123-133
• /
• 1996

Recently, a greater interest has been placed on the methods of acration control for management of maloders during composting. This study was performed to evaluate the influence of the intermittent aeration method on the composting temperature and chemical components. Combinatoins of cattle manure and rice hulls were placed on a batch composter(3,150-L size) practically designed. The chemical properties of the produced compost were suitable for plant growth, when composting temperature was maintained in the range from 45 to $66^{\circ}C$ for 42 days. It should be noted that cooling effect of intake air was not significant because air temperature was as high as $30^{\circ}C$. Heat and aeration loss from the composter did not affect biomass decomposition by microbes during composting. Mixtures of materials with C/N ratios higher than 30 required longer composting period.

• 대한산업공학회지
• /
• 제34권4호
• /
• pp.481-488
• /
• 2008

This paper deals with a scheduling problem for two-machine flow shop, in which the preceding machine is a batch processing machine that can process a number of jobs simultaneously. To minimize makespan of the system, we present a mixed integer linear programming formulation for the problem, and using this formulation, it is shown that an optimal solution for small problem can be obtained by a commercial optimization software. However, since the problem is NP-hard and the size of a real problem is very large, we propose a number of heuristic algorithms including genetic algorithm to solve practical big-sized problems in a reasonable computational time. To verify performances of the algorithms, we compare them with lower bound for the problem. From the results of these computational experiments, some of the heuristic algorithms show very good performances for the problem.

• 한국환경과학회지
• /
• 제31권4호
• /
• pp.295-309
• /
• 2022

The purpose of this study was to evaluate the operational characteristics of wastewater treatment using Sequencing Batch Reactor (SBR) with Aerobic Granular Sludge (AGS) separator in the pilot plant. Pilot plant experiments were conducted using SBR with AGS separator and pollution removal efficiencies were evaluated based on the operational condition and surface properties of AGS. The results of the operation on water quality of the effluent showed that the average concentration of total organic carbon, suspended solids, nitrogen, and phosphorus was 6.89 mg/L, 7.33 mg/L, 7.33 mg/L, and 0.2 mg/L, respectively. All these concentrations complied the effluent standard in Korea. The concentration of mixed liquor suspended solid (MLSS) fluctuated, but the AGS/MLSS ratio was constant at 86.5±1.3%. Although the AGS/MLSS ratio was constant, sludge volume index improved. These results suggested that the particle discharged fine sludge and increased the AGS praticle size in the AGS. Optical microscopy revealed the presence of dense AGS at the end of the operation, and particles of > 0.6 mm were found. Compared to those of belt-type AGS separator, the required area and power consumption of the hydrocyclone-type AGS separator were reduced by 27.5% and 83.8%, respectively.