There are several really good books on orchestration. Samuel Adler's book sits next to my workstation and I refer to it often as a reference. But for me, the best instruction comes from the study of how composers orchestrate their own music. Maurice Ravel, who is celebrated as an orchestrator, wrote several pieces first for piano before realizing them for orchestra. One can see how certain pianistic effects are translated into orchestral texture and can also learn how music which appears difficult in piano form is actually a simpler harmonic idea written with energy for the piano (tremolos for example). Ravel also orchestrated the music of others and a study of his problem-solving techniques (translating keyboard into orchestral) is priceless.
Composers also 'reverse engineer' their music - writing something first for orchestra and then making a piano version. For those that don't play piano, this is a valuable teaching tool.
I will admit here to my basic nature as an editor. I didn't always like to write because my inclination is to jump right to the finished product. However, as soon as I understood that I'm a better editor than a writer, writing actually became fun. My first step is always to get the ideas written down as fast as I can and then edit the hell out of them. Same thing for orchestrating music.
In 1986, I started playing with a good, established quintet, taking over for an exiting tuba player. That group had in its repertoire Gustav Mahler's Wo die schonen Trompetten blasen from Das Knaben Wunderhorn by another arranger. But it was full of holes and inconsistencies, and I set out to edit the arrangement. I became so frustrated trying to clean up the other arrangement that I eventually went back to Mahler's original version for voice and piano and made a new arrangement. It turned out pretty well in spite of my thin experience at the time.
Obviously, a quintet has only five voices and opportunities to play with color are limited by that. Arranging for quintet is often about seeking compromises which 'imply' musical intentions rather than accurately portraying them. Compromise is not a bad word.
Later, when writing for the large brass group, I revisited that same Mahler song and reveled in the voices and colors available, as well as several years of experience writing for brass in various combinations. Looking at arranging more Mahler pieces for the Boulder Brass, I first studied how Mahler re-crafted the Wunderhorn songs from the original piano & voice versions (sometimes contemporaneously) for his symphonies. Similarly, he orchestrated Songs of a Wayfarer several years after his original voice and piano version, and careful study reveals Mahler's own approach to coloring his own music. To be clear, I made no attempt to transcribe the Wunderhorn and Wayfarer songs from the orchestral versions, but was comparing the orchestral versions to the piano versions to learn how Mahler approached problem solving when orchestrating his own music. I did borrow some techniques and some of his percussion writing from the orchestral scores.
I have a two-screen setup for my workstation and when working on a piece like this, I will put the PDF of both the keyboard and orchestral version side by side on the screen to my right, and the Finale score on which I'm working on the screen directly in front of me. This way, I have instant access to the orchestration techniques of the composer. It can be difficult to manage this on the screen, but worth the effort. One could also do this with a smaller screen but I would suggest putting just the orchestral version on the screen with the hardcopy of the keyboard version on your desk.
Superficially, orchestration is about assigning parts and solving problems. It becomes fun (and artistic?) when instrumentation and problem solving are just the first steps - that moment when you take the step from being a mere technician, to the manipulation of musical colors with intent.