In multi-operation forging processes, the process fault due to missing parts from dies is a critical concern. The objective of this paper is to develop an effective method for detecting missing parts by using automatic classification of tonnage signals during continuous production. In this paper, a new feature selection and hierarchical classification method is developed to improve the classification performance for multiclass faults. In the development of the methodology, the signal segmentation is conducted at the first step based on an offline station-by-station test in a forging process. Afterwards, the principal component analysis is conducted on the segmented tonnage signals to generate the principal component (PC) features to be selected for designing the classifier. Finally, the optimal selection of PC features is integrated with the design of a hierarchical classifier by using the criterion of minimizing the probabilities of misclassification among classes. A case study using a real-world forging process is provided in the paper, which demonstrates the effectiveness of the developed methodology for detecting and diagnosing the missing parts faults in the multiple forging operation process. The classifier performance is also validated through the cross-validations to achieve a given average classification error.