The current fault detection in industrial plants is basically dependent on the operators’ skill and experience even though automatic fault handling is strongly preferred in view of complexity of contemporary plants and the heavy workload of operators thereby. There seems to be two difficulties in developing such handling system. One is the poor capability of analysis algorithm or tool itself, and another is the intuitive approach of operators, which is difficult to be understood or formulated. In this paper, we introduce a prototype analysis software based on wavelet transforms which we are developing for the purpose of fault handling. Wr describe the analysis software in detail first, and then, explain some preliminary results obtained from actual data collected in our oil refinery.