Control Chart Patterns Recognition Based on Optimized Deep Belief Neural Network and Data Information Enhancement
Control chart patterns (CCPs) are often used for quality control in the manufacturing process, and effective recognition of these patterns is critical to manufacturing. In the dynamic production process, the raw data and features of CCPs are used to recognize or further predict the trends. However,...
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | article |
| Lenguaje: | EN |
| Publicado: |
IEEE
2020
|
| Materias: | |
| Acceso en línea: | https://doaj.org/article/7c7bfb02ab8e48ee87084f21fca551ab |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Sumario: | Control chart patterns (CCPs) are often used for quality control in the manufacturing process, and effective recognition of these patterns is critical to manufacturing. In the dynamic production process, the raw data and features of CCPs are used to recognize or further predict the trends. However, the inaccuracy of CCPs information extraction, loss of information, and complex recognizer can lead to the difficulty of recognition. In order to improve the accuracy of information extraction and recognition, a CCPs recognition method based on optimized deep belief network (DBN) and data information enhancement was proposed. Adaptive features selection and information enhancement (AFIE) was used to select the most appropriate features and make these features combine with the raw data to from the dataset in order to reduce the data dimension, and then combine dimensioned data with the selected features to enhance the data information. Further, this study presented a DBN with three restricted Boltzmann machine structures, which was optimized by using the artificial fish swarm algorithm (AFSA). The method of AFIE was discussed to obtain the optimal data set, and parameters of the network structure were analyzed, optimized, and discussed based on experiments and AFSA. At the same time, this method was compared with multi-layer perceptron neural network. The simulation results showed that the method proposed in this study exhibited excellent effect, and the recognition accuracy achieved by this method was 99.78% for 2000 samples of each pattern. |
|---|