Numerical simulation is widely used in product design for different purposes ranging from structural modelling, FEM analysis, vibroacoustics to multi-objective design optimisation. Unfortunately, its application to Sound Quality is still extremely limited. This paper describes the use of a multi-objective optimisation code aimed at identifying noise control solutions which can improve the sound quality at the operator station of earth moving machinery during real working conditions. In a previous study the case of stationary noise signals was analysed and a multi-objective genetic algorithm was used to find the modifications in the input spectrum which led to the minimization of the time-averaged values of loudness and sharpness. In this paper the optimisation algorithm was modified to be applied to time-variant noise signals, characteristic of real working conditions. New input variables were identified to describe the time variant characteristics of the input signals and a numerical code was developed according to the DIN-45631/A1 procedure in order to properly calculate the loudness parameter of time-variant sounds. The new multi-objective genetic algorithm was finally applied to different noise signals recorded at the operator position of loaders in working conditions, with the purpose to find the modifications in the input system which minimised the percentile values of loudness and sharpness parameters. The results confirm the significant link between sound quality condition and frequency content of the noise signals, making it possible to evaluate the spectral variations needed to obtain psychoacoustic improvements.