Of all the works of the Dwarves, none can rival
Khazad-dûm in vastness, grandeur, or in later years horror.  It was initially settled by Durin's Folk, and the population later swelled when refugees from Gabil-gathol and Tumunzahar arrived after the destruction of their cities at the end of the First Age.  The realm continued its glorious ascent until the Third Age until the Dwarves dug too deep into the earth, looking for mithril, and released the Balrog.  Khazad-dûm, the "Dwarf-mansion", was abandoned and went from halls of light and slendour to a pit of darkness: Moria.

The first half of the name is Khazad and means "Dwarf, Dwarves". Tolkien says that Khazâd is "form in composition khăzăd-". Technically then Khazad is plural. This is the only word that actually shows the difference between the "normal" form, which Tolkien never specifically names, and the "composition form".  We can see that the <â> is shortened to <a>.  The composition form here matches how the construct state is formed in Hebrew. Here, in the compound, it is used in the plural form in an attributive manner, as opposed to the singular khuzd. This leads to the translation "Mansion of the Dwarves", since Khazad-dûm was home to an entire civilization.  If translated as an English compound, we would say "Dwarf-mansion", as in English saying "Dwarves-mansion" is awkward and doesn't sound right since we often use singulars in compounds.

Thus Khazad is "dwarves" and is plural, nominative, composition.

Tolkien states that "dûm is probably a plural or collective = excavation(s), hall(s), mansion(s)' ".  Note the use of "(s)", indicating an optional plural interpretation.  If dûm is a collective, it would be a bit odd to translate into English since these words have only singular and plural forms, not a collective.  I would say it's likely that dûm is collective because the form fits into the pattern of singulars, as per CuCC in Khuzd.  The root of dûm would probably be either a geminated/doubled root, D-M-M, or some kind of weak root, D-Y-M or D-W-M.  If weak, I would lean towards guessing the latter.  In Khuzdul, <y> is apparently changed to <i> when at the end of a syllable, as per aya
ai- "upon".  It is then very possible that <w> changes to <u> in the same circumstances.  Arabic has mandatory phonetic changes when these consonants are between certain vowel combinations, leading to a collapse to a single vowel or diphthong.  Having <w> as the weak radical would make it more likely that the resulting vowel would be <û>.   Geminated/doubled roots, as D-M-M, regularly drop the final consonant and lengthen the vowel when placed in a vocalization that has a single vowel, as in dumm → dûm.

The contact between the two elements has <d-d>, which forces us to consider whether one is actually a <t>.  Felak-gundu is sometimes written as Felag-gundu, showing voicing assimilation, where the first consonant is assimilated to the second.  Arabic has the same process with -dt-
-dd-We see Khazâd in the battle cry Baruk Khazâd! Khazâd ai-mênu!  That <d> seems stable.  If the <d> in dûm is actually <t>, it might very well be assimilated to <d>, even though this would be the reverse direction from the examples from Khuzdul and Arabic.  If true, this would mean dûm is originally tûm before assimilation.  This actually leads to a very elegant explanation of both dûm (tûm) and Tumunzahar.  Here, I will end the discussion on dûm by saying that I believe the unassimilated form is tûm, and that it is simply a different inflection (in number) of tumun- in Tumunzahar.  See the entry on Tumunzahar for more information.

Dûm then is
a voice-assimilated version of tûm meaning "excavation(s), hall(s), mansion(s)" and is collective, nominative, indefinite.

Khazad-dûm, as a whole, can be translated as "Mansions of the Dwarves", "Dwarf-mansion", "Dwarf-delving", etc., and is a compound word with noun-noun word order.

The Lord of the Rings, The Fellowship of the Ring, Book II, ch 3
The Silmarillion, index, pg 325, 337
Parma Eldalamberon XVII: Words, Phrases and Passages, pg 35
The Lost Road, pg 274