Early detection of process abnormality is an important issue for manufacturing industry, particularly for an industry which is predominantly using manual operation. The result of early detection could reduce manufacturing cost due to unsatisfactory product quality. This paper reports a detection of process abnormality in food extruder using principle component analysis. To demonstrate the effectiveness of the procedure, an experiment is designed and performed on a pilot plant where we simulate the abnormality that might occur in an industrial production plant. Specifically, we target the variations of moisture content in feed materials. The experimental data are analyzed for detection of abnormality in the product.