Rule based generation of random data

I'm fiddling with a personal project that can fill a universe with random data. I am looking for a generic system that can suggest combinations of features for some entity at random but affected by rules.

So for example. I want to be able to generate planets with features "size, distance from star, temperature, atmosphere, water content". Some of those features do not affect each other. Distance from star is inversely related to temperature, which in turn affects atmosphere and water presence.

I want to do something similar with life forms, plants, biomes, individuals.

My first attempt I created a list of features and a system where I could say, this goes with that at a certain probability. So no atmosphere means water is very unlikely. A very small planet means atmosphere is impossible or thin at best. It does kind of work, but the system is inflexible and requires several pages of rules often for silly things like if a planet is small it cannot be large. So I'm looking for something better. I googled, but I don't really know the proper words to use ;)

My question to you is: Do you know algorithms by name or any method or source I can investigate that will help me solve this problem?