With the development of intelligent manufacturing, there is a growing demand for Chinese-named entity recognition (CNER) techniques for identifying the terms and concepts from manufacturing process texts to lay the foundation for knowledge extraction and management. Manufacturing process texts have special characteristics such as concept complexity and diversity, contextual ambiguity, and lack of standardization, which makes it difficult for the existing CNER method to identify entities from them accurately. This paper presents a hybrid neural network model to identify the entities from Chinese manufacturing process texts to address the above issues. The proposed model uses a multi-granularity dynamic word vector generation method to learn the correlation between components in Chinese process texts and incorporates external knowledge into the model, which can improve semantic expression and better understand contextual information. Then, the Transformer and IDCNN (Iterated Dilated Convolutional Neural Network) are integrated to annotate the data sequences, where the Transformer model solves the long-distance dependency problem, and the IDCNN focuses on the local information, combining the features of the two networks, thus improving the recognition accuracy. Finally, CRF is used to learn the labeling dependencies and filter out the correct labeled sequences. The experiment results show that the proposed model performs better than the existing CNER models (BERT-BiLSTM-CRF and its variants), improving 7%-17% in the F1 score. Manufacturing-specific named entities.