The article discusses the need for a transformer design software program that can be easily used by beginners in the field without the need for trial and error. Most small-sized transformer manufacturers in South Korea do not use such software due to the high price, instead relying on spreadsheets like Excel. The present study aims to overcome the problems of conventional methods by finding a design method that does not require expert knowledge and obtaining all possible designs to avoid missing optimal values. The developed software program is introduced and its usefulness is demonstrated experimentally in various chapters of the paper.

The article discusses the need for a transformer design software program that can be easily used by beginners in the field without the need for trial and error. Most small-sized transformer manufacturers in South Korea do not use such software due to the high price, instead relying on spreadsheets like Excel. The present study aims to overcome the problems of conventional methods by finding a design method that does not require expert knowledge and obtaining all possible designs to avoid missing optimal values. The developed software program is introduced and its usefulness is demonstrated experimentally in various chapters of the paper.

The efficiency of transformers is critical to save electrical energy. There are three types of current transformer design methods. The current methods have problems. In this paper, we described current ways of designing a transformer, and we also explained their weakness. To overcome these weakness, we apply deep learning to designing a transformer. This paper also shows that our model recommend design parameters to satisfy design specification.

The ways of present transformer design can be divided into two types: using spreadsheet like Excel and using design software. Our first work is to develop the design software that can be used by small companies for free. We developed design software which generates as many designs as possible, without aid of expert's knowledge. Our second work is to improve efficiency by parallelizing the repetitive calculations. The performance of our design software is evaluated by comparing the results obtained from different numbers of cores. We show improvement of execution time where execution of 4 cores is about 45% faster than that of 1 core.

본 논문에서는 스마트 센서 응용 소프트웨어를 테스팅하기 위한 효율적인 방법인 가상 센서 시스템을 제안한다. 센서 응용 소프트웨어를 테스트하는 보편적인 방법은 테스트 보드에 센서를 직접 연결시켜 테스팅 환경의 센서 측정값으로 응용소프트웨어를 테스팅 하는 것이다. 센서 측정값을 입력으로 센서 응용 소프트웨어를 테스팅함으로 발생하는 문제는 테스트 데이터가 제한적이라는 것이다. 즉, 테스트 데이터가 센서로부터 생성되기 때문에 소프트웨어 테스터가 테스트 데이터를 조절하지 못하는 문제가 있다. 이러한 문제를 해결하기 위해 가상센서 시스템을 제안한다. 가상 센서 시스템은 소프트웨어 테스터가 센서의 측정값을 조절할 수 있게 한다. 가상 센서 시스템에서 센서 선택, 센서 특성화, 출력 패턴 정의의 세 단계를 통해 가상 센서를 정의한다. 가상 센서 시스템을 통해 조절 가능한 센서 측정값을 사용함으로써 센서 응용 소프트웨어에 대한 효율적인 테스트가 가능하다. 본 연구의 유용성을 보이기 위해 가상 센서 시스템을 안드로이드 앱의 센서 프로그램에 적용해 보고 실험 결과를 보인다. This paper presents a virtual sensor system that is an effective method to test application software of smart sensors. The common way of testing sensor application is to build a test board, connect sensors to the board, and test sensor applications on the board with sensor's measurements as inputs. The problem of testing sensor application software with sensor's measurements as inputs is the restriction of test data. In other words, software testers cannot manipulate test data, because test data is generated by sensors. To solve this problem a virtual sensor system is presented in this paper. The virtual sensor system enables software testers to manipulate measurements of sensors. In the virtual sensor system, generation of virtual sensors comprises three stages - sensor selection, sensor characterization, and determination of output patterns. Sensor's measurements that can be manipulated through the virtual sensor system make the process of testing efficient. To show the usefulness of our virtual sensor system, it is applied to sensor applications in Android platform and the result of experiments is shown.

Errors misusing union member are expressions that refer the value of a union variable with different type from the type of the value. A program with errors misusing union member may occur runtime errors or may return inappropriate results. An analyzer for detecting errors misusing union member is required because the existing compilers in C/C++ does not handle those. In this paper we propose an idea that detects those errors. The proposed method detects errors misusing union member using type annotation and type expressions that compute the types of union members.

This paper addresses a problem of violating the member accessibility of a class in C++, which is not detected as an error by existing C++ compilers. The member access specifiers can be used to specify member accessibility. The C++ uses a private or protected specifier for specifying the members which cannot be accessed from outside of an object. However, the private or protected members can be accessed from outside of that object by the pointer arithmetic in C++. We show some violating examples that cannot be detected by existing C++ compilers. The contribution of this paper is to discover and define the new problem of the member accessibility.