Abstract
The automaton theory has been widely applied in many areas of science, such as system simulations, neural networks, etc. It is also a basic theory of compiling routine for discriminating words of a programming language. This paper analyses the equivalence theorem between finite automata and regular expressions, points out the problem existing in the construction rules from finite automata to regular expressions, and proposes further a construction approach from finite automata including multi-node loops to regular expressions. The usage of this theorem and the construction process of regular expressions are expatriated, and the principle of choosing loops and their simplification from finite automata to regular expressions are discussed in detail