the software to be developed will be given by a fixed-model and a data-mining technique able to identify global patterns in the observed data. The integration of these two methods allows data-driven changes in the structure of the model, without the drawbacks of a totally direct-feedback model. Another advantage of the hybrid approach is that model adaptability can also take into account deviations due to non-measurable events, such as components aging, environmental effects, or combustion deposits.