There is such constant changes but it's not at all irratioinal. Have you noticed that the hiragana for bi is just the hiragana for hi plus two dots? (び and ひ) And similarly, the hiragana for ba is just the hiragana for ha plus two dots. (ば and は) This is pretty similar to the changes, for example, the negative prefix in English - irregular, impossible, inefficient, illegal, etc. Why it should be ir-, im-, in-, il-...? It could just be because of a tendency to simplify pronunciation. I see that you know Chinese, and here is an example in Chinese. When two syllables with the 3rd tone are combined to create a word, the first syllable will be read as 2nd tone. Thus, 莞尔 [wǎn ěr] is actually read as [wán ěr]. Such a change also simplifies pronunciation. In Japanese, you actually see such changes between 清音(せいおん), their correspondent 濁音(だくおん) and 半濁音(はんだくおん). If you consider the writing of the hiraganas, such changes are really quite rational~

Mark has already explained it, but if you want details: It's called Rendaku (連濁), voicing of the following consonant. See http://en.wikipedia.org/wiki/Rendaku. So it's actually 花 "hana" and 火 "hi". The other process that can be irratiting for learners is Sokuon (促音), consonant lengthening: Trailing -tsu, -chi and sometimes -ku of the first Kanji is merged with the following non-voiced consonant of the second Kanji into a single long (geminate) consonant. So 出 shutsu + 席 seki = 出席 shusseki, 学 gaku + 校 kou = 学校 gakkou, and so on.

Thanks for your help :D Now I understand it!!