Fault detection using hierarchical self-organizing map

The appropriate features are essential for pattern classification and signal modeling. Stamping operations eagerly need the condition monitoring system in practice to guarantee the product quality; however, its processes are nearly intractable and the features of its signals are not easy selected. The self-organizing map (SOM) is an excellent tool in data exploratory due to its property of mapping the complex relationships in high dimensional space onto simple geometric relationships in a low dimensional space. A hierarchical SOM was developed in the paper: the prototype vectors of the bottom layer SOM are considered as the features, which are clustered at the top layer SOMs. The results demonstrate that the proposed approach working effectively in the condition monitoring. This suggests that the hierarchical SOM is worthy of more applications.