Forecasting the abnormal operations in batch units and limiting their incidence have been the emphasis on the study of the on-line monitoring batch operation. The abundant real-time data gathered in the automatic system contain the complexity and uncertainty of the process behavior, so mining and fusing a series of universal applicable quantities from the historical data to monitor the operation status is essential. In this paper, a combination of wavelet analysis (WT) and the fractal encoding (FE) technique is proposed to resolve the problem by exploiting their virtues. Using WT with a local-time frequency analysis, this strategy takes advantage of the multiresolution representations on the measured profiles. Adapted FE is used to segment regions within the multiresolution representations. By extracting fractal models, the proposed method, like the philosophy of traditional statistical process control, can generate simple monitoring charts, track the progress in each batch run, and monitor the occurrence of observable upsets. Due to the local property of FE, the on-line batch monitoring based on the proposed method of filling the missing values only at some local regions is good enough for detecting the current status. Additionally, the advantages of the proposed method are demonstrated through two sets of benchmark data, a DuPont industrial batch polymerization reactor and fed-batch penicillin production, which are characterized by some fault sources.