A typical method for rule reduction of a PID fuzzy controller is to divide the three-term into two separate PD and PI parts. A further reduction is possible if the controller is switched from PD to PI-type after a certain period of time. In that case only a single set of rules will be executed at a time and thus the controller rule base will be reduced. A further simplification is possible if a single rule-base is used for both the PD and PI-type FLC. This means that the fuzzy sets for change of error and sum of error will be redefined within the same universe of discourse, i.e., the fuzzy sets for both change of error and sum of error will be the same. Such a strategy is adopted in this paper. Accordingly the fuzzy sets are restored by tuning the scaling factors for change of error and sum of error using a single neuron network with non-linear activation function. Genetic algorithms are, on the other hand, used to train the neural network. The proposed method is tested and validated in the control of a single-link flexible manipulator.