In the semiconductor manufacturing process of etching, temperature control is a very important factor. The stability of the etching rate and quality of uniformity are related to accuracy of temperature control. In situation that there are some defects of deigning machine, process of gas changes in vacuum chamber is one of reasons that cause the temperature becomes unstable. To improve the problem on the premise that changes the structure of machine lowest is the purpose of this search, and therefore we develop a smart temperature controller for raising the stability of temperature in process and improving the yield. We used the learning ability of neruo-fuzzy to optimize the status of every membership functions. That will let the temperature controller achieve the optimum effect of control during pressure be changed.