Deep reinforcement learning based dynamic scheduling decision-making framework

이승훈 Graduate School, Yonsei University 2021 국내박사

Effectiveness of wastewater treatment under a sustainable supply chain management

Sao, Sreymouy Graduate School, Yonsei University 2024 국내석사

요즘 사람들이 가장 중요하게 구매하는 상품 중 하나는 옷인데, 한 사람이 1 년에 평균적으로 구매하는 옷의 수가 엄청나게 증가했다. 섬유 제품은 사회의 기본적인 필요를 충족시키는 데 중요한 역할을 한다. 섬유는 다양한 용도로 사용될 수 있지만 가장 인기 있는 것은 옷과 용기이다. 수건, 카펫, 덮개가 있는 가구, 창문 덮개, 냅킨, 침대 시트 및 기타 평평한 표면 덮개는 모두 가정에서 사용되는 섬유 사용의 형태들이다. 섬유 제품은 천연 또는 합성 섬유, 필라멘트, 실 및 원사로 만들어진다. 우리는 일반적으로 섬유를 우리가 입는 옷으로 생각한다. 현대 사람들의 세계에서 섬유 제품은 날이 갈수록 인기가 많아지고 있다. 따라서 섬유는 태어나서 죽을 때까지 우리 삶의 모든 면에서 중요한 역할을 한다. 이 때문에 수익이 발생하여 경제 성장으로 이어진다. 이는 섬유 산업이 다양한 국가의 경제에서 중요한 축이 되었고, 소규모 및 대규모 기업을 대표한다는 것을 의한다. "섬유 산업"은 전세계적으로 엄청난 경제적 영향을 미치고 있다. 경제 성장을 도왔음에도 불구하고, "섬유 산업"은 기후 변화에 강한 영향을 미치는 환경 오염에 책임이 있다. 섬유 산업에서 발생하는 오염은 환경에 중대한 부정적인 영향을 미치며 그 원인은 매우 분명하다. 섬유 산업에서 생산되는 "폐수"는 매일 수백만 갤런의 물을 사용하여 표백, 염색, 섬유 세척 후 완성된 제품을 세척한다. 그 산업이 엄청난 양의 섬유 제품을 생산할 때, 습식 가공에서 나오는 폐수뿐만 아니라 엄청난 양의 탄소 배출이 발생하고, 이는 사회와 환경을 해롭게 한다. 섬유 생산 과정에서 낭비되는 물의 양이 많고, 이러한 폐기물은 수중 생물에 대한 독성 영향의 원인이 됩니다. 포름알데히드, 염소, 중금속 등의 물질을 일상생활에서 섭취하는 것이 일반적이다. 이 물질들은 물속에 버려진다. 섬유는 일상생활에서 필수품이기 때문에 많은 제품이 생산되고 적절한 관리가 필요하다. 섬유산업 관리에서는 폐기물 관리를 해야 한다. 그렇지 않으면 오염과 정부의 규제로 인해 회사가 국가에서 생존할 수 없다. 이러한 이유로 우리는 자외선과 가시광원하에서 광촉매 공정이라는 첨단 공정을 도입하여 폐수처리와 함께 섬유산업을 적절하게 관리하기 위해 본 연구를 개발하게 되었다. 이 공정은 화학반응을 이용하여 폐수를 처리하는 가장 간단하고 효과적인 방법이다. 이 공정을 활용하여 섬유산업의 폐수를 정화하는 연구는 어떤 문헌에서도 수행되지 않았다. 고전적 최적화는 Karush-Kuhn-Tucker 조건(KKT)과 함께 의사결정변수에 대한 글로벌 최소 해결점을 찾는 데 도움이 된다. 숙련된 노동비용의 적용은 본 연구의 민감도 분석에서 알 수 있듯이 대부분의 SCM 에서 민감한 파라미터일 수밖에 없다. 마지막으로, 자외선 (UV) 및 가시광원하에서 광촉매 공정을 통해 폐수를 처리하기 위해 가변적인 생산율과 투자를 적용하면 섬유 공급망 시스템이 비용 효율적이고 환경 친화적이라는 결론을 얻을 수 있다. One of the most important products that people purchase nowadays is clothes, and the average number of clothing items that one person purchases in a year has massively increased. Textile products play a vital role in the basic needs of society. Textiles can be used for a variety of things, but the most popular ones are for clothes and containers. Towels, carpeting, upholstered furniture, window coverings, napkins, bed sheets, and other flat surface coverings are all forms of textile use in the home. Textile items are made from fibers, filaments, thread, and yarn, either natural or synthetic. We generally consider textiles to be the clothes we wear. In the modern world of people, textile products are becoming more and more popular day by day. Therefore, textiles play a significant role in all facets of our lives from birth to death. For this reason, revenue is generated, leading to economic growth. This implies that the textile industry has become an important pillar in the economies of a variety of countries and represents a diverse range of enterprises, both small-scale and large-scale. The “textile industry” has a massive economic impact around the world. Despite helping economic growth, the “textile industry” is responsible for environmental pollution with a strong impact on climate change. The pollution that the textile sector produces has a significant negative influence on the environment, and the causes are quite clear. Every day, the “wastewater” produced by the textile industry utilizes millions of gallons of water to clean the completed product after bleaching, dyeing, and washing the fiber. When the industry produces a huge amount of textile products, a huge amount of carbon emissions as well as wastewater from the wet processing are generated, which are harmful to society and the environment. During the textile production process, there are high amounts of water that are wasted, and these wastes are responsible for the toxic effects on aquatic life. It is commonplace for people to consume substances such as formaldehyde, chlorine, and heavy metals during their daily activities. These substances are dumped into bodies of water. Since textiles are a necessity for daily life, many products are produced, and proper management is required. During the management of the textile industry, the manager needs to manage the waste; otherwise, the company cannot survive in the country due to the pollution and the regulations from the government. For this reason, we developed this research study to manage the textile industry properly along with wastewater treatment by introducing an advanced process called photo-catalytic processes under ultraviolet (UV) and visible light sources. This process is the simplest and most effective way of using chemical reactions to treat wastewater. Contrary to the literature, no research was conducted on utilizing this process to purify wastewater for the textile industry. The classical optimization, along with Karush-Kuhn-Tucker conditions (KKT), helps find the global minimum solution for decision variables. The application of skilled labor costs is inevitably a sensitive parameter in most SCM, as shown by the sensitivity analysis in this present study. Finally, it can be concluded that applying variable production rates and investment to treat wastewater through photo-catalytic processes under ultraviolet (UV) and visible light sources makes the textile supply chain system cost-effective and environment-friendly.

Analysis of turnover factors of employees by job field in the Biopharmaceutical industry

Park, Jungtae Sungkyunkwan University 2022 국내박사

연구배경: 미래 신성장동력산업으로 관심이 확대되고 있는 바이오제약 산업은 현재 업계의 급격한 발전으로 인해 인력난을 겪고 있으며, 인력 부족의 주된 원인 중의 하나로 종사자들의 잦은 이직이 확인되었다. 특히 이직 발생률이 높은 분야는 생산, 영업/마케팅, 그리고 임상/RA 분야로 이들 3개 분야의 이직의도에 대한 상세하고 포괄적인 이해는 한국의 바이오제약산업 성장에 매우 중요하다고 볼 수 있다. 연구목적: 본 연구는 바이오제약 산업 성장의 핵심 요소인 생산, 영업/마케팅, 임상/RA 분야별 종사자들의 정확한 이직요인을 파악하고 그에 대한 대책을 마련하여 기업의 이직률을 줄이고 기업이 안정적으로 성장을 이룰 기회를 생성, 궁극적으로 바이오제약 산업의 발전에 기반이 되는 틀을 마련하는 것을 그 목적으로 한다. 연구방법: 구글 설문조사 방법을 이용하여 바이오의약품 회사의 생산, 영업/마케팅, 임상/RA 분야에서 일하는 종사자들에 대한 이직요인을 조사하였다. 설문조사는 2020년 9월부터 2020년 10월까지 약 2개월 동안 진행하였고, 설문 대상자는 인구통계학적 변수에 대한 quota를 기준으로 선정하여 국내외 바이오의약품 업체 종사자들을 대상으로 500명을 조사하였다. 설문지 구성은 개인특성관련요인, 직무관련요인, 조직관련요인 등을 고려하여 설정하였으며, 이직의도와 각 요인의 상관관계를 확인하기 위해 다중회귀분석을 실시하였다. 연구결과: 직종별 주요 이직요인으로 생산직 종사자는 조직몰입도 및 감독자에 대한 만족도가 높을수록, 그리고 부양가족의 책임이 낮을수록 이직의도가 낮은 것으로 조사되었다. 영업/마케팅 종사자는 여성보다 남성이 이직의도가 낮고, 또한 중견/대기업에 다니고 있거나, 직무자체 및 봉급에 대한 만족도가 높을수록 이직의도가 낮은 것으로 조사되었다. 임상/RA 종사자는 중견/대기업 종사자일수록, 그리고 직무자체에 대한 만족도가 높을수록 이직의도가 낮은 것으로 확인되었다. 결론: 본 논문에서 바이오제약산업의 직무, 조직, 개인문제를 결합한 다양한 상황에서 3개 직종별 이직요인을 분석하였으며, 연구결과에서 생산,영업/마케팅.임상/RA 각 직종별 이직요인 분석에 차이가 있는 것으로 나타났다, 특히 생산직종은 조직몰입도 및 감독자 그리고 가족부양 책임부분이 이직요인으로 분석되어 영업/마케팅. 임상/RA 의 주요 이직요인인 중견/대기업 근무여부, 직무자체의 만족도 와는 다르게 분석되었으며, 또한 영업/마케팅은 추가적으로 성별에 따라, 봉급 만족도에 따라 이직요인에 차이가 있는 것으로 분석 되었다 이와같이 3개 분야별 이직요인 분석에 차이가 있기 때문에 이직율을 줄이기 위해서 각 직종별 종사자들의 이직의도 요인 해소를 위한 needs를 파악하고 차별적인 방안수립이 필요하다. 본 연구의 결과를 바탕으로 바이오제약 기업에서 발생하는 이직의 주된 요인을 확인하고 이직률을 줄일 수 있는 적절한 방안을 수립하여, 기업이 안정적으로 성장할 수 있는 기틀을 제공할 수 있을 것으로 기대한다. Background: The biopharmaceutical industry is rapidly growing due to persistent and ever-increasing demand for medical products and an aging population. The rapid growth of the biopharmaceutical industry inevitably demands an increase in the highly skilled workforce. Recruiting and maintaining professional workers is mandatory in three areas that dictate the growth and success of an organization, namely production, sales/marketing, and clinical/RA. However, these fields have the highest turnover rate among other fields in biopharmaceutical industry and thus a detailed and comprehensive understanding of turnover intentions in these fields is very important for the growth of the Korean biopharmaceutical industry. Purpose: This work aims to contribute to the growth and development of the domestic pharmaceutical industry by suggesting correct improvement measures and strategies for employees in the fields of production, sales/marketing, and clinical/RA, to reduce unnecessary turnover and helping the companies to optimize their HR management Method: Google survey was used to find the turnover factors of employees working in three fields. The survey was conducted from September 2020 to October 2020, and 500 survey participants were selected from domestic and foreign biopharmaceutical companies, based on their demographic variables. The composition of the questionnaire was set in consideration of turnover intention factors related to individual characteristics, job-related factors, and organization-related factors. Multiple regression analysis was performed to confirm the significant correlation between turnover intention and each turnover intention factor. Results: As a major turnover factor, it was found that the higher the level of organizational commitment and satisfaction with supervisors, and the lower the family obligation, the lower the turnover intention for production employees. As for sales/marketing employees, it was found that males had lower turnover intentions than females, and employees in mid/large sized companies or with higher satisfaction with their job itself and salary also had low turnover intention. For clinical/RA, employees in mid/large sized companies or with higher satisfaction with their job itself found to have lower turnover intention. Conclusion: In this study, the turnover factors for the three job fields have been analyzed in various conditions combined with job-related, organizational-related, and personal characteristics in the biopharmaceutical industry. It was found that there were significant differences in the turnover factors of each field. In particular, the turnover intention for production employees found to be lower when they have organization commitment or no family support obligation, or satisfy with their supervisor, whereas the turnover intention for sales/marketing and clinical/RA found to be lower when they work in mid/large size companies or when their job satisfactions are met. In addition, the intentions were also affected by gender or salary for sales/marketing employees. It is expected that the companies will identify the turnover factors and exact needs of employees derived in this study and use such findings to establish a plan to leverage job satisfaction in accordance with internal conditions and compensation policies of a corporation, thereby providing a foundation for stable growth of the company.

(The) dynamic impact of low-cost carriers on South Korean tourism and aviation industries : application of Lotka-Volterra mòdel

Khan, Nokhaiz Tariq Sungkyunkwan University 2019 국내박사

Lowered travel expense is one of the key drivers in promoting tourism. The emergence of low-cost carriers (LCCs) in the aviation industry has dramatically decreased the cost of short-to-medium distance air travel, alongside the growth of the tourism industry. LCCs have also changed the competitive dynamics within the aviation industry, challenging the traditional market pre-dominated by the conventional full-service carriers (FSCs). Past studies have demonstrated the scale of impact of LCCs on the aviation market, but have yet to explore the intricate dynamics of competition in greater detail. This study practices the Lotka-Volterra model along with the development of the concept of moving-window analysis to assess the changing impact of the entrance of LCCs on South Korean tourism and airline industries. The results of this study conclude that the impact of LCCs over tourism industry is in line with the literature and LCCs have aided the positive tourism growth by inducing their own new market. The impact of LCCs on aviation market is not static rather it is dynamic. The relationship between LCCs and FSCs evolved over time. The dynamics of LCC-FSC competition depend upon the market size and the market concentration. Furthermore, in response to the emergence of South Korean LCCs in aviation market, FSCs launched their own LCCs. This concept known as airline-within-airline (AWAs) or carrier-within-carrier (CWCs) has laid the possibilities for the fight back of FSCs against the aggressive penetration of LCCs into the aviation market. The Results offer insights to industry players on developing sustainable strategic plans for the future routes and tourism destinations in Asia and beyond.

Taxi service pricing model for public interest purposes based on demand and supply analysis and forecasting using spatial temporal self-attention mechanism

김상윤 Graduate School, Yonsei University 2020 국내석사

교통은 한 사람의 인생 사이클과 경제적 가치 사슬에 영향을 미치기 때문에 사회의 필수적인 요소 중 하나이다. 택시업계도 대중교통의 일원으로서 적지 않은 역할을 하고 있다. 그러나 현재 대중교통 체계가 다양해지고 우버 등 새로운 교통 플랫폼이 등장하고 있어 택시업계의 문제점이 부각되고 있다. 택시운전자가 난폭운전과 불법 주정차등으로 인한 교통문제에 많이 연루되어 있음은 잘 알려져 있다. 반대로, 공공의 덕목의 측면으로 택시는 특히 도시와 교외 지역에서 가장 다재 다능한 교통수단이다. 택시 요금 모델의 재구성은 대중 교통으로서 택시 산업의 덕목을 유지하기 위한 후보 접근법이 될 것이라 기대한다. 이번 연구에서는 2014년 뉴욕의 택시 옐로우 캡 사용 기록 데이터베이스를 수요와 공급 예측의 측면으로 분석했다. 우리는 STSAN을 활용해 미래 택시의 수요와 공급을 예측됐다. 예측한 택시 사용 기록 데이터베이스는 뉴욕시의 시간과 공간에 따른 수요와 공급 패턴의 측면으로 분석됐다. 택시 사업의 경우 특정 시기와 지역에서 공급과 수요가 균형을 이루지 못했다. 공급의 수요는 특히 맨해튼에 편향되어 있었다. 따라서, 택시 사업 이익의 불안정을 해소하기 위해, 예측된 수급 균형을 이용하여 후보 가격 책정 모델을 구축했다. 채택된 가격 모델은 낮은 수익을 내는 대표 예시의 이익을 증가시킴으로써 시뮬레이션에서 운전자 사이의 이익 격차를 줄일 수 있었다. 수요와 공급 균형에 대한 추정을 통해 택시 산업의 공공성을 극대화하기 위한 가격 책정 모델을 제시할 수 있었다. 후보 모델은 이익 불균형과 불안정성을 낮추었다. 택시기사의 수입 불균형과 불안정을 해소하는 등의 직접적인 문제를 해결함으로 인해 난폭운전과 교통혼잡 등 각종 사회적 비용을 줄일 수 있을 것이라 추측한다. 우리는 이 연구의 가격결정 모델이 사양산업으로 분류되는 택시업계의 구성원들이 수익구조의 개선을 통해 삶의 질을 향상시킴으로써 변화하는 사회에 적응할 수 있는 환경을 제공할 수 있다고 주장할 수 있다. Transportation is one of the essential elements of society because it affects people’s lives as well as the economic value chain. The taxi industry is a part of the public transportation system. Currently, because public transportation systems are varied and new transportation platforms such as Uber have emerged, the taxi industry has encountered severe problems. Reckless driving and illegal parking are well-known traffic problems caused by taxis. In terms of public transportation, taxis are the most versatile mode of transportation, especially in both urban and suburban areas. The reconstruction of a taxi pricing model will be beneficial for maintaining the characteristic of taxis as public transportation. In this study, we analyzed a taxi movement database for demand and supply prediction. We classified the New York City Taxi and Limousine Commission (TLC) Yellow cap taxi data based on zip codes using reverse geocoding. The flow of future taxis was predicted using the spatial temporal self-attention network. The predicted taxi movement database was analyzed in terms of demand and supply patterns according to the time and region of New York City. Based on the distribution patterns of demand, supply, and demand/supply ratio by the time–space dimension, the supply and demand were not harmonized at specific times and regions. The distribution of supply and matched demand were highly biased to a certain region, particularly in Manhattan. Even though the taxi business is one of the on-demand businesses affected by customer demand, the fare system did not reflect the demand of customers. Therefore, for resolving the instability of taxi business profits, we constructed a candidate pricing model using a predicted supply/demand balance. The adopted pricing model decreased the profit gap between drivers in simulations by increasing the profit of the lower revenue case. In addition, the suggested model indicated less variations in terms of revenue for the various circumstances that we analyzed. By estimating the demand and supply balance, we suggest a pricing model for maximizing the public aspect of the taxi industry. The candidate model elevated profit imbalances and instability. It can reduce various social costs, including eliminating the disproportionality and instability of the revenue of the taxi driver. We claim that the pricing model in this study provides an environment where members in the declining taxi industry can adapt to changing societies by enhancing their quality of life by improving social virtues and benefits.

Development of IoT-based sensors, big data processing, and prediction model for real-time monitoring system in manufacturing industry

Syafrudin, Muhammad Dongguk University 2019 국내박사

With the increase in the amount of data captured during the manufacturing process, monitoring systems are becoming important factors in decision making for management. Current technologies such as Internet of Things (IoT)-based sensors can be considered a solution to provide efficient monitoring of the manufacturing process. In this study, a real-time monitoring system that utilizes IoT-based sensors, big data processing and a prediction model is proposed. Firstly, an IoT-based sensor that collects temperature, humidity, accelerometer and gyroscope data was developed. The characteristics of IoT-generated sensor data from the manufacturing process are as follows: real-time, large amounts and unstructured type. The proposed big data processing platform utilizes Apache Kafka as a message queue, Apache Storm as a real-time processing engine and MongoDB to store the sensor data from the manufacturing process. Secondly, for the proposed prediction model, Density-Based Spatial Clustering of Applications with Noise (DBSCAN)-based outlier detection and Random Forest classification were used to remove outlier sensor data and provide fault detection during the manufacturing process, respectively.The proposed model was evaluated and tested at an automotive manufacturing assembly line in Korea. The results showed that IoT-based sensors and the proposed big data processing system are sufficiently efficient to monitor the manufacturing process. Furthermore, the proposed prediction model has better fault prediction accuracy than other models given the sensor data as input. The proposed system is expected to support management by improving decision-making and will help prevent unexpected losses caused by faults during the manufacturing process. In addition, the overall proposed framework can be used as practical guidelines for the industrial practitioner in order to adopt IoT-based sensors, big data processing and machine learning model in manufacturing industry.

Data-driven inventory management modeling using entropy-regularized deep reinforcement learning

정지원 Graduate School, Yonsei University 2023 국내박사

효과적인 재고 관리 시스템은 기업 성공 좌우에 무시못할 영향을 끼치게 됩니다. 의류 산업과에도 소매점의 효과적인 재고 관리는 기업이 높은 보관 비용이 발생하지 않도록 적절하고 충분한 재고량으로 품목이 고갈되지 않고 안전한 보관 수준을 유지하는 과정으로 경쟁력 및 고객 충성도를 유지하는 핵심 성공 요인입니다. 그러나 불행하게도 경쟁력 있는 재고 관리 시스템은 도전 없이는 이루기가 어렵습니다. 4차 산업혁명의 도래와 함께 자율적으로 비실용적인 산업 문제를 해결하기 위해 인공지능 전문성의 필수가 높아졌습니다. 인공지능 연구는 기계 학습을 기반으로 하며 그 중에 강화학습도 포함됩니다. 변수의 차원 및 문제의 규모가 점점 현실화됨에 따라 심층강화학습이 등장했습니다. 연구 단계의 초시점에서는 Atari 게임을 시뮬레이션하기 위해 많이 사용하게 됐습니다. Atari 게임에서 고성능 결과의 달성을 보여 본 게임은 점진적으로 개발된 최첨단 심층 강화 학습 모델 개발에 프레임워크가 되어줬습니다. 본 연구에서는 Soft Ac-tor-Critic에 중점을 두고 의류 Make-to-stock 재고 관리 문제를 해결하기 위해 심층강화학습 방법에 대한 연구를 수행했고 또한, 비용 부분에서 효율적인 재고 관리를 더욱 촉진하기 위해 Total Penalty로 지명된 새로운 보상 함수를 제안합니다. 본 연구에서 사용된 알고리즘은 재고 관리의 가장 중요한 핵심 성과 지표 중 일부인 Service Level, Sell-through Rate 및 Inventory-to-sales ratio 기반으로 서로 평가됩니다. Soft Actor-Critic 이 가장 낮은 총 비용으로 재고량의 흐름을 유지하고 수요를 충족하는 데 있어서 제일 뛰어난 성능을 보였습니다. 핵심 성과 지표 측면에서 Soft Actor-Critic은 과잉 재고 없이 판매할 수 있는 재고의 가용성을 높임으로써 서비스 수준과 판매율 간의 균형을 개선하면서 재고 관리 시스템을 운영했습니다. Soft Actor-Critic은 Twin Delayed Deep Deterministic Policy Gradient(총 비용을 보상 함수로 학습) 및 (R, Q) – Policy (총 페널티를 보상 함수로 학습) 보다 2.42% 및 31.91% 더 낮은 총 비용을 달성했습니다. 또한 SAC는 (R, Q) – Policy 보다 81.39% 낮은 재고 대비 매출 비율을 달성하여 재고 재고를 가용한 상태로 판매하고 비용을 최적화하는 더 우수한 능력을 보여줬습니다. The success of enterprises heavily relies on their inventory management systems, which largely determine their competitiveness and customer loyalty. This case is also relevant in the garment industries. The primary goal of establishing a successful inventory management system is to maintain safe storage level of items without risking depletion, thereby avoiding excessive storage costs. Unfortunately, achieving a competitive inventory management system poses challenges. With the emergence advent of the Fourth Industrial Revolution, expertise in Artificial Intelligence has become essential for addressing complex industrial problems autonomously. The study of Artificial Intelligence is founded on Machine Learning, and Reinforcement Learning is part of it. As the dimensions of variables and complexity of problems increase, Deep Reinforcement Learning has emerged by combining Deep Learning and Reinforcement Learning. Its initial prominence was seen in the simulation of Atari games, where it served as framework for developing state-of-the-art Deep Reinforcement Learning models by achieving high-performance results. This study focuses on applying Deep Reinforcement Learning methods to address an apparel Make-to-stock inventory management focusing specially on Soft Actor-Critic. Additionally, a novel reward function called Total Penalty is proposed to enhance cost-efficient inventory management. The models are evaluated based on critical Key Performance Indicators, including sell-through rate, service level, and inventory-to-sales ratio. Soft Actor-Critic, in both scenarios when trained with Total Cost and Total Penalty as reward functions, demonstrated superior inventory management performance by meeting demands at the lowest Total Cost. Furthermore, Soft Actor-Critic achieved a favorable balance between service level and sell-through rate within large span by ensuring optimal stock availability without excessive overstocking. Compared to Twin Delayed Deep Deterministic Policy Gradient (with Total Cost as reward function), and (R, Q) – policy (with Total Penalty as reward function) models, Soft Actor-Critic achieved a 2.42% and 31.91% lower Total Cost, respectively. Additionally, Soft Actor-Critic achieved an 81.39% lower inventory-to-sales ratio than (R, Q) – policy, indicating its superior ability to optimize costs and make inventory stocks available for sales.

(An) integrated framework of energy simulation and PLM for sustainable manufacturing

Zhao, Wenbin Sungkyunkwan university 2018 국내박사

The world has undergone rapid development in the economy, science, industry, and other fields since the Industrial Revolution while also grappling with various social and environmental problems. Humanity’s dependence on fossil fuels for daily living has been a major cause of global warming as it has generated massive amounts of greenhouse gases. The indefinite material development and unbridled competition under the market economy have caused a great deal of social dilemma in terms of human dignity while also destroying the earth’s fragile ecosystem. People have begun to understand that the unfettered exploitation and consumption of natural resources that has been supporting the modern human civilization is no longer able to sustain human prosperity and civilization, as it has become obvious that the linear production and consumption based on fossil fuels has clear limitations. In this phase, keywords such as “low-carbon green growth,” “clean production,” and “eco-friendly products” began to be adopted by the manufacturing industry, which had already been grappling with industry-specific problems, along with the increasing influence of energy prices among the various manufacturing costs. It is necessary to consider the reduction of energy use early in the engineering stage, such as in the product design and production stages. Now is also the right time to expand the application of simulation, an engineering front-loading methodology. As a result, the energy simulation methodology, which has not been widely utilized thus far, has come to be actively discussed of late. The product lifecycle management (PLM) solution is composed mainly of the product data management (PDM) system and various other applications, while it is also necessary to manage additional information for a host of applications, ranging from the PLM system for sustainable manufacturing to the existing PDM system for sustainable development. To carry out such energy simulations, it is necessary to expand the information capacity and enhance the data management and utilization capability compared with the existing PLM, and to further expand it into a PLM system capable of supporting sustainable manufacturing. In this paper, a PLM information model for sustainable engineering and an integrated framework for simulation/optimization for energy saving purposes were designed, implemented, and then verified. The results of this study showed that manufacturing enterprises would be able to contribute to sustainable development for the human society by reducing their energy consumption and generation of environmental pollutants while at the same time improving their productivity and product quality and reducing their manufacturing cost.

Machine learning-based process monitoring and maintenance for smart manufacturing

이슬기 Korea University 2018 국내박사

Current manufacturing industries accumulate large amounts of process data due to advances in sensing, communication technology, data storage, and use of standards. The challenge has shifted from collecting a huge amount of process data to analyzing these data for decision making. Manufacturing companies have tried to analyze process data to produce high-quality product and to adapt quickly to changing conditions. In such situations, multivariate statistical process control (MSPC) has become a key technology for process data analytics for better quality of products, stability, and operation efficiency enhancement in modern manufacturing industries. The main objective of MSPC is to quickly detect the occurrence of assignable cause signal, so that correct action may take place before quality and system degrades and faulty units are produced. Although conventional MSPC techniques can be successfully applied to simple manufacturing processes, these techniques are unable to handle the large streams of complicated process data originated from the nonnormal, high-dimensional, nonlinear, and time-varying situations found in modern manufacturing. To confront these challenges, several machine learning algorithm-based MSPC techniques have been developed. Machine learning algorithms facilitate the discovery of useful information from sufficient amounts of data. The main aim of this dissertation is the integration of machine learning algorithm and MSPC technique to effectively handle the various problems encountered in modern manufacturing. The proposed monitoring methods in this dissertation use a representative machine learning algorithm, such as support vector machine and deep neural network, which can take into consideration the nature of the process data. Especially, the proposed methods are capable of handling important issues in complex manufacturing processes: (1) nonnormality, (2) high-dimensionality, (3) nonlinear relationship between variables, and (4) time-varying characteristics. To address both nonnormal and time-varying issue, this dissertation proposes the time-adaptive support vector data description-based control chart. Because the proposed chart can promptly reflect the time-varying conditions, large number of false alarms are reduced and stability of manufacturing system can be maintained. In addition, this dissertation proposes variational autoencoder (VAE)-based monitoring method to reduce any misdetections. This resolution is due to the VAE’s capability of simultaneously handling nonnormality, nonlinearity and high-dimensionality. In this dissertation, experiments with simulation and real data from manufacturing process were conducted to compare various existing models. The results demonstrated that the proposed monitoring models were superior to existing model in terms of reducing false alarms and misdetections. I believe that all of the proposed methodologies in this dissertation promising in that they can be a solution for real problems in modern manufacturing, and thus enable smart manufacturing.

Blockchain technology for supply chain management : a holistic approach for adoption, implication, and challenges

Javed, Aslam Sungkyunkwan University 2023 국내박사

Blockchain is a technology that helps the organization uplift firm performance by improving supply chain (SC) integration, agility, and security through real-time information sharing, end-to-end visibility, transparency, data management, immutability, auditability, irrevocable information, and cyber-security platforms. The oil and gas (O&G) sectors are negatable to researchers, and there are limited studies on O&G supply chain management (SCM). Additionally, there is no empirical evidence that suggests implementation of blockchain for O&G as a solution for potential problems. This study has made an initial effort towards proposing a framework that shows the problems and challenges for the oil and gas supply chain under its segments (upstream, midstream, and downstream) and provides the interlink among blockchain properties for SCM problems. Furthermore, this study analyzes the impact of SC integration, SC flexibility, and financial security on firm performance in a blockchain environment and understands the robustness capabilities as a mediator between SC parameters and firm performance. SC managers were selected for survey questionnaires from the Pakistan Oil and Gas industries. The perception of SC managers was analyzed, and the result revealed that the SC manager believes that the blockchain enables SC to increase firm performance because blockchain technology is reflected as high-tech to support the firm process, responses, methods, and its help to eliminate bottlenecks, avoid uncertainties, and improve decision making, which leads to improved firm performance and robustness capabilities. This study guides managers about the potential problems of existing SC and how blockchain solves SC problems more effectively. 블록체인은 실시간 정보 공유, 엔드투엔드 가시성, 투명성, 데이터 관리, 불변성, 감사 가능성, 취소 불가능한 정보, 사이버 보안 플랫폼 등을 통해 공급망 (SC) 통합, 민첩성, 보안을 개선해 조직의 성과 상승을 돕는 기술이다. 하지만 기존의 연구자들은 석유 및 가스 (O&G) 분야에 블록체인 기술을 적용하는 것에 대해 부정적이었으며 O&G 공급망 관리 (SCM)에서의 제한된 연구만이 수행되었고, 블록체인 기술이 O&G에서 발생하는 문제들을 해결할 수 있다는 실험적인 연구는 수행된 적이 없다. 본 연구는 O&G 공급망의 세그먼트 (업스트림, 미드스트림 및 다운스트림)에 따른 문제점들을 보여주고 SCM 문제에 대한 블록체인 속성 간의 상호 연계를 제공하는 프레임워크를 제안하기 위해 초기 노력을 기울였다. 또한 본 연구는 블록체인 기술이 적용된 환경에서 SC 통합, SC 유연성 및 재무 보안이 기업 성과에 미치는 영향을 분석하고 SC 매개변수와 조직 성과 사이의 매개자로서 갖는 강건성을 설명한다. 본 연구의 유용성을 입증하기 위해 파키스탄 O&G산업의 SC 관리자들을 선정하여 인터뷰를 수행하였다. SC 관리자들의 응답을 통해 블록체인 기술이 SC의 병목현상 제거, 불확실성 회피를 통해 확실한 절차, 대응, 방법을 통한 의사결정에 도움이 되는 첨단 기술로써 SC의 성과와 강건성을 높일 수 있음을 확인하였다. 본 연구는 기존 SC에서의 잠재적 문제를 블록체인 기술을 통해 어떻게 효과적으로 해결하는지를 관리자에게 설명한다.