Axial fans are widely used in heavy machines due to their ability to produce high flow rate for cooling of engines. At the same time, the noise generated by these fans causes one of the most serious problems. This work is concerned with the low noise technique of discrete frequency noise. To calculate the unsteady resultant force over the fan blade in an unsymmetric engine room. Time-Marching Free-Wake Method is used. From the calculations of unsteady force on fan blades, noise signal of an engine cooling fan is calculated by using an acoustic similarity law. Noise optimization is obtained from Neural Network which is constructed based on the calculated flow rate and noise spectrum.