1994 年 110 巻 6 号 p. 479-486
This paper describes the use of an artificial neural network for the data analysis of the plasma synthesis of acetylene from coal and other hydrocarbons using 100kW plasma reactor. The experimental data have wide variety of experimental conditions such as electric power input, raw material species and their feed rate, quenching conditions, and so on. 89 experimental data sets were stored in a 4-layered neural network having 37 input units and 6 output units within the maximum error of 7.4%. The results were compared with those of linear regression. Although the extrapolation was not always satisfactory, the interpolated data estimated by the trained neural network were found to be excellent. It is expected that neural network technique can be a useful and versatile tool for the data analysis as well as statistical techniques.