Industrial plant noise is the result of a combination of many various sound sources. The spectral composition of these sound sources varies depending of the equipments: wide-band noises mainly related to cooling towers, tonal noises such as transformers and some other noises with spectral components in middle frequencies. The aim of this paper is to investigate the pleasantness/unpleasantness of industrial noise with a “sound design” approach. Sixteen different industrial stimuli were created combining cooling tower noise, transformer noise and another industrial source noise, mixed with two kind of background noise (“traffic road” and “nature”). With a computer tool, twenty-two people were able to change the composition of the industrial sound (without changing the overall SPL of 40 dB(A)) and design a better sound target by improving sound quality. The first result of this study is that the average optimized stimulus is composed with more of cooling tower noise and less transformer noise. This average stimulus is “a little more pleasant” than the original sound (on a continuous scale from “much more unpleasant” to “much more pleasant”). This increase in pleasantness (with constant overall level) has been translated into an equivalent reduction of about 2.5 dB(A) of the original sound.