Data Mining Method for Offshore Structures based on Big Data Technology

다국어 초록 (Multilingual Abstract)

As many products as ships and offshore structures are constructed in the shipyard, and various data are generated and stored in the design or construction stage. Big data technology needs to be applied to process data of large size quickly, obtain meaningful results and use it for decision making. In this paper, we propose a solution to two of the problems that may occur in the shipyard.
One of the two problems which can arise in the shipyard has mainly happened in the design stage. Engineers can make the mistake of choosing the wrong material in the design process, and the wrong material selection in the design process can directly lead to a design error. Another problem may arise during the procurement and purchase process. In the absence of additional information such as lead time of material or inventory at the time of procurement, additional time is required to retrieve the data. Both problems arise predominantly from the unskilled. Therefore, the purpose of this study is to establish a
8
system that can inform the engineers about the relationships between materials which can be obtained by association analysis and material requirements which can be obtained by regression analysis. This kind of system can help the engineers to reduce design errors and time consuming due to the procurement process.
The information of piping materials used in an offshore structure can be regarded as ‘big data’ because of their various types and size, and the data mining algorithms based on the big data technology are applied to data related to the offshore structures. To analyze the relationship between materials for design, ‘frequent pattern growth algorithm’ was used. For material requirement analysis, big data technology-based regression analysis was used to generate a regression model, respectively.
Finally, the proposed method was used to check the relationship between materials, and to predict material requirement, and verified the effectiveness of the proposed method by comparing each result with actual cases.

국문 초록 (Abstract)

조선소에 많은 선박과 해양 구조물들이 건조되면서 다양한 데이터들이 설계 및 건조 과정에서 생성되고 누적된다. 누적된 데이터를 빠르게 처리하여 의사설정에 이용하려는 필요성에 발생함에 따라 빅데이터 기술의 도입 필요성도 함께 커지고 있다. 본 연구에서는 조선소에서 발생할 수 있는 두 가지 사례에 대하여 빅데이터 기반 데이터 마이닝 방법을 통한 해결책을 제안하고자 한다.
첫 번째 문제점은 설계 단계에서 발생할 수 있는 문제로, 설계 과정에서 적절하지 못한 자재를 선정하여 그것이 오작으로 이어지는 경우이다. 또 다른 한가지는 구매 및 조달 과정에서 발생할 수 있는 문제로, 조달 과정을 관리하기 위한 자재 관련 추가적인 정보를 검색하는데 추가적인 시수가 소요된다는 점이다. 두 가지 문제 모두 미숙련자에게서 주로 발생하며, 본 연구에서는 연관성 분석을 이용한 자재 추천과 회귀 분석을 이용한 소요량 예측이라는
90

목차 (Table of Contents)

• Contents
• Abstract 7
• 1. Introduction 9
• 1.1. Research background 9
• 1.2. Related works 14
• 1.3. Target of the study 15
• 1.4. System configuration 17
• 2. Data mining method for pipng material 19
• 2.1. Piping materials of an offshore structure 19
• 2.2. Association analysis 23
• 2.2.1. Assotication between piping material for recommendation of associated materials 23
• 2.2.2. Comparison of association analysis algorithms 25
• 2.3. Regression analysis 31
• 2.3.1. Assistance of purchase process by forecasting material requirement 32
• 2.3.2. Regression model training 35
• 2.4. Estabilishment of big data framework 46
• 2.4.1. Big data framework 46
• 2.4.2. Data processing 47
• 3. Application to big data framework 49
• 3.1. Recommendation of associated materials 49
• 3.1.1. Method of association analysis of piping material 49
• 3.1.2. Result of analysis and validation 56
• 3.2. Forecasting of material requirement 63
• 3.2.1. Method of training regression model 63
• 3.2.2. Prediction of material requirement and validation 69
• 4. Conclusion and future works 85
• Related Works 86
• 국문 초록 89