The anomalies in product manufacturing process are related to product defects, and the accurate detection of these anomalies is conducive to improving product quality. The feature-based control chart pattern recognition (CCPR) method has been widely applied to this problem. However, most of the existing feature methods only focus on the amplitude characteristics of the data, ignoring the structural characteristics and sequence relations of the data. A novel feature extraction and recognition method of control chart pattern (CCP) based on multi-delay weighted ordinal pattern (MDWOP) is proposed. MDWOP features integrate the amplitude and sequence structure characteristics of the data, and comprehensively characterize the complexity of CCP from different scales based on time delay parameters. An ensemble classifier recognition method based on multi-delay features is proposed to improve model recognition accuracy. Simulation results show that the average accuracy of the proposed method for eight small fluctuation CCPs is 95.44%, and is better than that of the single classifier method.