To address the limitations of traditional machining parameter optimization methods in adapting to dynamic machining conditions and ensuring high precision, this study proposes a deep learning-driven adaptive optimization framework for CNC milling parameters. First, a hybrid CNN-LSTM model integrated with Bayesian optimization (BO) is established to predict key machining outputs, including surface roughness (Ra), cutting force (Fc), and material removal rate (MRR). The model leverages convolutional neural network (CNN) to extract spatial features from multi-sensor signals (vibration, temperature, current) and long short term memory (LSTM) to capture temporal dependencies in the machining process. Second, a multi-objective adaptive optimization model considering machining precision, efficiency, and energy consumption is constructed, with constraints on cutting force, power, and tool wear. The model is solved by an improved particle swarm optimization (PSO) algorithm embedded with real-time process feedback. Experiments are conducted on a 5-axis CNC milling machine using aluminum alloy AA6061 and titanium alloy Ti-6Al-4V, with Taguchi experimental design and extended full factorial design for data collection. Results show that the proposed CNN-LSTM-BO model achieves superior prediction accuracy compared to single CNN, LSTM, and traditional regression models. The adaptive optimization framework reduces Ra by 18.7% − 23.5%, improves MRR by 12.3% −16.8%, and lowers specific energy consumption by 10.2% −14.6% compared to empirical parameters and non-adaptive optimization methods. This research provides a data-driven intelligent solution for high-precision CNC milling parameter optimization under dynamic working conditions.
Xiaoli Qu (Sun,) studied this question.