Abstract: In the flotation process it is hard to develop an accurate mathematical model for both concentrate grade and tailing grade which characterize the key technical indices. However, it is also hard to control efficiently the process by conventional methods. A new way is therefore recommended combining the case-based reasoning (CBR) with conventional control methods, i.e. the intelligently optimal index setting by CBR. In this way the experience of flotation process can be grasped more profoundly from lots of historical process data so as to build a case database which summaries typical operation conditions. The retrieval, reuse, revision and store of those cases are discussed in CBR. The intelligently optimal index setting model can update automatically all setpoints in every control loop so as to avoid the subjectivity and randomness due to arbitrary manual control setting. This approach has been successfully applied to a flotation process in a mineral processing plant with its effectiveness actually proved.