PURPOSE: To ensure constantly stable supercharging characteristics and to improve torque performance, by a method wherein a feedback control region is learnt according to a deviation between the supercharging pressure of intake air and a desired supercharging pressure, and from the learning result, a present feedback control region is discriminated.

CONSTITUTION: A correction amount computing means (c) is provided for computing the proportional part of a feedback correction amount, and the integral part thereof by which a deviation between a supercharging pressure detected by a supercharging pressure means (b) and a given desired supercharging pressure is decreased to zero, and a discriminating means (d) is provided for learning a feedback control region based on the deviation, and selecting a present feedback control region from the learning result, and selection the proportional part and the integrating part of a feedback correction amount according to the discriminating result. Based on an output from an operating state detecting means (a), a fundamental value for controlling a supercharging pressure is computed by a control means (e), correction is effected according to the selected feedback correction amount, and the result is outputted as a supercharging pressure control signal to a supercharging pressure control means (f).