Evolutionary operation (EVOP) is a continuous improvement system which explores a region of process operating conditions by deliberately creating some systematic changes to the process variable levels without jeopardizing the product. It is aimed at securing a satisfactory operating condition in full-scale manufacturing processes, which is generally different from that obtained in laboratory or pilot plant experiments. Information on how to improve the process is generated from a simple experimental design. Traditional EVOP procedures are established on the assumption that the variance of the response variable should be small and stable in the region of the process operation. However, it is often the case that process noises have an influence on the stability of the process. This process instability is due to many factors such as raw materials, ambient temperature, and equipment wear. Therefore, process variables should be optimized continuously not only to meet the target value but also to keep the variance of the response variables as low as possible. We propose a scheme to achieve robust process improvement. As a process performance measure, we adopted the mean square error (MSE) of the replicate response values on a specific operating condition, and used the Kruskal-Wallis test to identify significant differences between the process operating conditions.