The X-fluorescence analysis technique was introduced to measure the trace elements in archeological ceramic fragment samples. Production areas of the samples were expected to be properly and intelligently identified according to the differences in both element’s types and contents in samples. Aiming at the difficulties in spectrum analysis of the multi-element co-existence samples and in low count rates

the method of artificial neural network (ANN) was adopted to learn and identify the X-fluorescence spectra of samples. The total number of samples is 48 from 8 provinces and 20 gathering areas. Two kinds of structures of ANN are introduced to train and identify ceramic fragments in two classes of area domain

respectively. The correct rate of recognition was up to 100% for the samples whose production areas were accurately classified

while the rate was more than 60% for others. The recognition results of the method are satisfying.