This paper has developed an on-line pattern recogniser to detect unnatural patterns on control charts. The recogniser can assist in the correction of assignable causes. The recogniser presented here is based on the grey relation model which is used to calculate the degree of relative similarity between observations routinely collected from X-bar charts and each unnatural pattern, such as a trend or cycle. According to the degree of relative similarity, the recogniser will identify whether an unnatural pattern exists in the data. To explain the design process and the evaluation of the performance of the recogniser, a set of commonly encountered unnatural patterns, such as trends, cycles, systematic, stratification, mixtures and sudden-shifts, was applied in this paper. The performance of the recogniser was evaluated by simulation. The results showed that the recogniser has effectiveness in detecting the following patterns trend, cycle, sudden-shift, systematic, and stratification.