Force sensor based online tool wear monitoring using distributed Gaussian ARTMAP network

Tool condition monitoring is paramount for guaranteeing the quality of the workpiece and improving the life time of the cutter. Force sensor has been proven to be one of the most effective means to depict the tool wear variation during the machining process. However, because of the disturbance of noisy signal and the complexity of tool wear topology, the feature vectors usually demonstrate the non-uniformly distributed shapes and complex category boundaries, which will deteriorate the classification accuracy greatly. In this paper, a distributed Gaussian ARTMAP (dGAM) network is presented to realize the condition classification of the tool wear process. The main characteristic of this method is that the distributed Gaussian probability density, instead of the hyper rectangle match function, is used to realize the mapping between the feature vectors and the committed node. Therefore, the classifier is insensitive to the noisy data and suitable for non-uniformly distributed data. Based on the dGAM model, a monitoring system was built to realize the incremental learning and online classification of tool wear states. The analysis and comparison with Fuzzy ARTMAP show that the proposed classifier is more accurate. This method casts some new lights on the tool wear condition monitoring in real industrial environment.