Heuristic rules(sensitizing rules) provide effective decisions for detecting nonrandom patterns on the control chart. To implement these rules sucessfully, it forces user to pay careful attention to sample point patterns. Furthermore, The ability to interpret a particular pattern requires experience and knowledge of the process. This paper provides neural network approach recognize the unnatural pattern on the control chart automatically as well as to heighten the sensitivity of the process change. The proposed method consists of Two-Step Neural Network Control Chart(NNCC) to perform the (원문참조)-R chart`s function simultaneously. The first step NNCC detects the change of the process by the traditional neural network approach. The second step NNCC detects the change of the process by learning heuristic rules. The proposed method`s performance is compared with traditional Shewhart (원문참조) - R chart in terms of type I & Ⅱ error. Two-step NNCC`s errors are calculated by the simulation result.