Analysis of Two-Phase Flow Pattern Identification Methodologies for Embedded Systems

作者：Franco E F Jr^*; Salgado R M; Ohishi T; Rosa E S; Mastelari N

来源：IEEE Latin America Transactions, 2018, 16(3): 718-727.

DOI：10.1109/tla.2018.8358647

• Two-Phase Flow
• Pattern Recognition
• Embedded System
• Neural Networks
• Data Processing

Main goal of Two Phase Flow Pattern Identification is to determine the correct pattern in terms of flow in a section of a pipe. Identification of these patterns has applications in several areas, such as chemical and oil industries. When it is intended to use the technologies of identification in real systems, several aspects must be considered. The objective of this paper is to investigate the possibility of embedded a methodology to identify flow patterns in devices that have limited processing power and memory capacity. In this sense, several statistical manipulations were made aiming to reduce the information needed to identify the patterns. Tests were conducted simulating reductions in the data size, without loss quality of the identification process. The hit rate was used to evaluate the results, obtaining with the best settings an average over 95% of accuracy. From the tests performed, we observed that it was possible to find an identification model efficient and computationally capable of being embedded devices.

• 出版日期2018-3