Application of Weighted Combination Model in Tool Wear State Identification

Tool Condition Monitoring (TCM) is an integral part of realizing intelligent manufacturing. In the field of TCM, a single model may be insensitive to a certain category, resulting in poor performance of the model. In order to identify tool wear state more accurately, a combination model based on weight distribution is proposed. The acoustic emission signals acquired in milling are taken as examples. The process of feature extraction is described. After constructing the data set, BP neural network, Elman neural network and SVM are used as sub models to identify the tool wear state. With the minimum error as the optimization objective, the weight of each sub model is assigned by PSO algorithm. The weighted combination model is established. The superiority of combination model is verified by the validation set, and its generalization performance is tested by the test set. The weighted combination model has been proved to be effective. As the components of this model can be replaced by different algorithms or sub models, it provides a new idea for the follow-up research.