Abstract: It is the most important to acquire knowledge in intelligent systems and it reflects the intelligence level of the system.To apply soft computingtools to acquire knowledge is a focus of researchers from all over the world.As a new soft computingtool,the rough set theory has enormous potential in applying and it has been widely applied in lots of fields such as pattern recognition,medical diagnosis,medical data analysis,image processing,quality control,fault diagnosis,data mining,process control and so on.In our detecting system,traditional methods could not satisfy requirements for the restricition of real conditions and factors of operators and intelligent methods must be adopted.In order to realize the intelligent detecting,how to dffectively acquire knowledge is the key problem.to apply the rough set theory to acquire a correlative elementary knowledge and then optimize the knowledge by neural networks can improve the efficiency of acquiring knowledge.The method of acquiring the correlative elementary knowledge with rough set theory in our intelligent detecting system is given in this paper.