Aiming at the problem of tool wear prediction under small samples, this study proposes a tool condition prediction model based on the Tent-ASO-BP neural network. Firstly, the collected vibration and cutting force signals are denoised, and time-domain, frequency-domain, and time-frequency-domain feature parameters are extracted using techniques like fast Fourier transform and wavelet packet decomposition. Subsequently, based on the principle that tool wear increases with the number of cutting passes, Pearson correlation analysis is applied to select feature parameters with a correlation coefficient of no less than 0.9, indicating a strong correlation with tool wear. Finally, the selected feature parameters are combined into a feature vector, which serves as input for training the Tent-ASO-BP neural network for tool condition prediction. Experimental results demonstrate that the combined approach of Pearson correlation analysis and Tent-ASO-BP neural network exhibits excellent learning capability, enabling effective prediction of tool wear in small sample scenarios. This study contributes to addressing the challenges of tool wear prediction in situations with limited data. By incorporating denoising techniques and extracting relevant feature parameters, the proposed model enhances the accuracy of tool wear prediction. The utilization of Pearson correlation analysis ensures the selection of highly correlated features, further improving the model's performance. The Tent-ASO-BP neural network demonstrates its potential as a reliable tool for predicting tool wear, making it suitable for practical applications. In summary, this study presents a tool condition prediction model based on the Tent-ASO-BP neural network and Pearson correlation analysis, specifically designed for small sample scenarios. The experimental results confirm the model's excellent learning capability and its effectiveness in accurately predicting tool wear under such conditions.