I am trying to undestand how PRAAT works and to recognize consonants through spectrograms. I Know that it's possible to distinguish fricative consonants from nasal ones (for example) but is it possible to distinguish their place of articulation? ie. /ħ/ and /h/ (Voiceless pharyngeal fricative and Voiceless glottal fricative)
It is possible, in principle, to distinguish consonant place of articulation with spectrograms, within certain limits of precision, but only given certain prior knowledge. It used to be a standard exercise that students would be given a spectrogram and they had to transcribe it.
The technique involves figuring out the formants of the consonant, usually via transitions with vowels. Any good introductory phonetics text that deals with acoustics will give you the main cues for distinguishing labial, alveolar and velar consonants in English. However, if you really want [ħ] versus [h], you have a bit of an uphill climb because most intro textbooks don't deal with exotica. It's not impossible: what you'd need to do is find acoustic studies of such consonants in the professional literature. Very roughly speaking, you'll find that you can best identify [ħ] from lowered F3. However, the letter [ħ] is not articulatorily or acoustically uniform across languages, this comparing Arabic and certain dialects of Somali is inappropriate because the two consonants are quite different, despite the transcriptional standards usually applied to the languages.
A good transitional source for going beyond basic English consonants is Ladefoged & Maddieson The sounds of the world's languages. The really best way to fill in the required knowledge is to obtain good illustrative recordings of the sounds in a specific language, and then try to see what measurable differences in formant frequencies there are between [t] and [ʈ], or [k] and [q].
Theoretically, if you have a really great catalog of zillouns of sounds of the world's languages, you might develop a program to assign one or more plausible descriptive labels such as "laminal alveolar", given just the acoustics. Nobody has done it yet.