Tool wear condition monitoring is vital for boosting machining efficiency and ensuring product quality. Traditional methods for tool wear condition monitoring have shown inadequate predictive accuracy and generalization when dealing with complex nonlinear relationships and temporal dependencies. To address these issues, this paper presents a novel tool wear condition monitoring model that integrates GBDT and GRU. The proposed model consists of three main components: a data preprocessing module, a GBDT (Gradient Boosting Decision Tree) feature selection module, and a GRU (Gated Recurrent Unit) training module. Given that tool wear is influenced by multiple factors, feature selection is performed using the GBDT method following signal acquisition and preprocessing. This step aims to identify key features that significantly impact tool wear, thereby reducing data dimensionality and enhancing the efficiency of the monitoring model. Subsequently, the GRU model is utilized to capture temporal dependencies and accurately identify the tool wear condition, which leads to improved monitoring accuracy and generalization ability. Experimental results from public and laboratory datasets have demonstrated the validity and generalization of the model, with training accuracies of 98.9% and 97.8% achieved.