Thermoluminescence (TL) dating using a regression method of saturation exponential in predose technique was performed for 39 porcelain samples from past dynasties of China. The sampling was made by a diamond core-drill and the core was cut into several 0.2mm thick pieces using a precision slicing machine. In the process of the drilling and slicing the sample was cooled with water. All TL measurements were performed by using a fully automated combined TL/OSL dating system (model Riso TL/OSL-DA-15). The testing dose and calibrating dose were applied by using the same beta irradiator mounted in the dating system, which is attachable to the sample changer for providing software controlled irradiation of sample in situ. The radioactive beta source is an Amersham 90Sr (plaque source type SIPK, 1.48GBq). A preparatory porcelain sample was used for a measurement procedure. 39 porcelain pieces were dated by this method. The results show that the ratio between the average TL ages and the archaeological ages was 1.10 and the standard deviation was 29%. The exponential regress method is more suitable for porcelain dating, it has two advantages over the standard pre-dose method (i.e. linear regression method). (1) Expanding measurable age range. (2) A good accuracy.