Abstract:TC4 titanium alloy is a representative α + β dual-phase titanium alloy, which is widely used in engine blades and other components of aerospace equipment. This material typically requires heat treatment for strengthening to optimize its performance. To ensure that the heat-treated parts achieve ideal performance, it is necessary to test their physicochemical properties after heat treatment, such as microstructure, hardness, and elemental composition. To develop a rapid in-situ detection technology for the physicochemical states of heat-treated metal parts, a classification model was proposed by combining laser-induced breakdown spectroscopy (LIBS) with independent component analysis (ICA) and a deep neural network (DNN). The microstructure, Vickers hardness, and spectral characteristics of TC4 titanium alloy samples with different aging grades were analyzed. The spectral signals were preprocessed using the ICA method, and the results were used to establish a DNN model. The classification performance of the model was verified and evaluated using indicators such as the confusion matrix. The results show that the microstructure of TC4 samples can be regulated through solid solution and aging treatment, and the mechanical properties change accordingly. The uneven distribution of sample elements during microstructure control and the difference in ablated mass caused by different sample hardness contributes to the distinguishability of LIBS spectra of TC4 samples. The established ICA-DNN model facilitates dimensionality reduction and sensitive feature extraction of the physicochemical properties from the spectral data. Comprehensive evaluation results indicate that the classification performance of the model is strong, demonstrating the feasibility of using LIBS to characterize the physicochemical state of heat-treated metal materials. Compared to traditional detection methods, LIBS technology, as an emerging frontier detection technology, offers significant potential for in-situ, real-time, and micro-loss quality monitoring of heat-treated materials in industrial applications.