Due to its weak rigidity, thin-walled parts are prone to chatter and deformation in the milling process, which aggravates tool wear. In order to improve the milling efficiency and surface quality of thin-walled parts, a tool wear state recognition method driven by the fusion of digital twin and support vector machine (SVM) is proposed. The feature vectors are extracted by time-frequency domain analysis and wavelet packet transform. The super parameters are optimized by grid search and cross validation (GSCV). Combined with SVM algorithm, the wear state recognition model of milling tool for thin-walled parts is constructed. The experimental results show that SVM algorithm has obvious advantages in the classification and recognition of high-dimensional and small sample data. The recognition accuracy of different milling cutter wear states reaches 96% and 90.16% respectively, and has good generalization ability. Combined with machine learning algorithm, a high fidelity and lightweight digital twin is constructed and embedded into the milling process monitoring platform of thin-walled parts, so as to solve the problems of real-time signal monitoring and online recognition of tool wear state in the machining process.