An unsupervised dual-regression domain adversarial adaption network for tool wear prediction in multi-working conditions

•An unsupervised DR-DAN for unlabeled multi-working conditions industrial data is proposed.•DR-DAN can realizes tool wear knowledge transfer between different working conditions.•Weight discrepancy restriction can enhance the local consistency between distributions.•Prediction consistency loss can match accurate degradation stages without label supervised.•DR-DAN can effectively predict tool wear in different milling machines experiment.