Soft Computing and Its Application (Aniati Murni Arymurthy)
SOFT COMPUTING AND ITS APPLICATIONS Aniati Murni Arymurthy*
ABSTRAK Tema lokakarya ini adalah Komputasi Sains dan Teknologi Nuklir. Topik yang dipilih untuk makalah ini adalah Komputasi Lunak (Soft Computing) dan Aplikasinya. Komputasi sains di bidang ilmu komputer dan teknologi informasi terbagi menjadi analisis numerik, komputasi lunak atau komputasi cerdas, dan metode formal. Ada juga yang memasukkan metode formal dalam komputasi cerdas. Komputasi sains atau ilmiah di bidang sains dan teknologi diterjemahkan sebagai disain dan analisis algoritma (langkah-langkah penyelesaian) untuk penyelesaian masalah matematik di bidang sains dan teknologi. Pentingnya komputasi ilmiah adalah dimungkinkannya simulasi terhadap fenomena alam dimana prototipe disain dan penyelesaiannya dapat dilakukan secara virtual sebelum secara fisik rancangan sistemnya dibangun. Komputasi lunak (soft computing) berbeda dari komputasi keras yang konvensional (hard computing) dalam aspek mampu mengakomodasi ketidak-pastian (uncertainty), ketidak-tepatan (imprecision), kebenaran yang bersifat sebagian (partial truth), dan aproksimasi. Hal tersebut mirip dengan model berpikirnya manusia. Makalah ini membahas logika fuzzy, jaringan syaraf tiruan, support vector machine, dan algoritma genetika. Beberapa aplikasi potensial di bidang nuklir, antara lain mulai dari rancangan reaktor nuklir intinya, pengendalian sistem dinamis operasionalnya, monitoring normal tidaknya fungsi reaktor, sampai ke optimasi resiko dan biaya disain dan pemeliharaan.
ABSTRACT The theme of this seminar is on Scientific Computation and Nuclear Technology area. The selected topic for this paper is on Soft Computing and Its Applications. Scientific computation in the area of computer science and information technology includes numerical analysis, soft computing or computational intelligence, and formal method. Sometimes computational intelligence covers formal method. The term of scientific computing covers the context of design and analysis of algorithms for numerically solving mathematical problems in science and engineering. The importance of scientific computing is to be able to simulate natural phenomena and to virtually prototypes engineering designs before they are actually built. Soft computing differs from conventional (hard) computing in the sense that, unlike hard computing, it is tolerant of imprecision, uncertainty, partial truth, and approximation. In effect, the role model for soft computing is the human mind. This paper discusses fuzzy logic, neural network, support vector machine, and genetic algorithm. Several potential applications in nuclear
*
Laboratory for Pattern Recognition and Image Processing -Faculty of Computer Science, University of Indonesia, UI Campus, Depok 16424, e-mail:
[email protected]
23
Risalah Lokakarya Komputasi dalam Sains dan Teknologi Nuklir 2010, Oktober 2010 (23-44 )
technology include nuclear reactor core design, nuclear reactor dynamic system controller, nuclear power monitoring and fault detection; risk and cost optimization in reactor design and maintenance.
I.
BACKGROUNDS
The theme of this seminar is on Scientific Computation and Nuclear Technology area. The selected topic for this paper is on the scientific computation side that is specifically related to computer science and information technology with applications among others related to nuclear technology. The term of scientific computing covers the context of design and analysis of algorithms for numerically solving mathematical problems in science and engineering. Traditionally, it is called numerical analysis. It deals with continuous quantities and considers the effects of approximations. The importance of scientific computing is to be able to simulate natural phenomena and to virtually prototypes engineering designs before they are actually built [1]. In a department of computer science like the Faculty of Computer Science, University of Indonesia, scientific computation has ever become an area (or a field) of interest besides the other fields, such as Information Systems, Computer Architecture and Real-Time Systems, and Software Engineering. The scientific computing field includes numerical analysis, soft computing (computational intelligence) and formal method [2]. Other Department of Computer Science may include formal method in computational intelligence term. Soft computing differs from conventional (hard) computing in the sense that, unlike hard computing, it is tolerant of imprecision, uncertainty, partial truth, and approximation. In effect, the role model for soft computing is the human mind [3]. The soft computing includes Fuzzy Logic, Neural Networks, Support Vector Machines, evolutionary computation like Genetic Algorithm, Machine Learning and Pattern Recognition. This paper discusses the topics under Soft Computing that will include fuzzy logic, neural network, support vector machine, genetic algorithm, in the context of machine learning and pattern recognition. The discussion will only presents the important basic concepts to present the ideas, related applications, and literature reviews on their applications in nuclear technology. As a result, this paper is organized as follows. Fuzzy Logic (FL), Neural Network (NN), Support Vector Machine (SVM), and Genetic Algorithm (GA) are presented in Section II to Section V, consecutively. The GA topic can also be found in other paper in this seminar, among others is written by Hilda Deborah et al. [4]. 24
Soft Computing and Its Application (Aniati Murni Arymurthy)
Furthermore, in the conventional management information and retrieval system, the information are usually retrieved based on metadata; while in a modern multimedia information retrieval system, the information is retrieved based on content [5]. The system is called Content Based Information (that could be in the form of text, image, audio, or multimedia) Retrieval System (CBIRS). The development of a CBIRS may utilize various soft computing methods. Several relevant applications will also be on CBIRS. Finally, this paper is closed by a summary and several potential research topics in Section VI.
II.
FUZZY LOGIC
II.1. Basic Concepts Standard set theory is difference from fuzzy logic set theory. Standard set theory says that the intersection of a set A and its complement (Ac) is a null set, while in fuzzy logic set theory it is not a null set. It can be explained by Fig. 1 [6].
(a)
(b)
(c) Figure 1. (a) In the standard set theory the air temperature is cool if it is between 50oF-70oF; otherwise it is not cool. (b) The temperatures belong to a fuzzy set only to some extent and belong to the set’s complement to some extent. (c) The 55oF has 50% degree of membership to ‘cool’ and 50% degree of membership to ‘not cool’. (Source: Kosko and Isaka [6])
25
Risalah Lokakarya Komputasi dalam Sains dan Teknologi Nuklir 2010, Oktober 2010 (23-44 )
In a standard set theory, a bivalent indicator function IA of a non-fuzzy subset A of X has the following value (Eq. 1). 1 if x is a member of subset A; IA(x)
= 0 if x is not a member of subset A.
(1)
The bivalent indicator function was extended to a multi-valued indicator or membership function as shown in Fig. 1(c).
II.2. Fuzzy Logic and Air Condition System Controller Suppose for an example, membership of the air temperatures can be grouped into three categories that can be illustrated as Fig. 2.
Figure 2. Groups of Air Temperatures Membership Fuzzy Function. Suppose an air conditioner system has a fuzzy controller with the following rules: • If the air temperature is cold (between 0oF-60oF), then the motor speed is slow = (0.6 * temperature + 4) rpm; • If the air temperature is cool (between 50oF-68oF), then the motor speed is medium = (0.5 * temperature + 10) rpm; • If the air temperature is hot ( => 55oF), then the motor speed is fast = 45 rpm. The question is, if the air temperature is 57oF, how fast is the motor speed? First of all, the graphics membership equation for each group of air temperature should be defined. The followings are an example of the graphics equation (Eq. 2) following the Sugeno-style fuzzy inference [7,8,9] for membership fuzzy function in Fig. 2. The equation for ’cold temperature’ group has a value of: 1 if temperature x