A digital twin-driven part spatio-temporal quality prediction framework integrated with equipment degradation state analysis

ABSTRACT

Key parts in high-value equipment have critical requirements of high precision and performance. The machining processes of such parts normally involve multiple stations. Therefore, the machining quality of finished parts is an accumulated result of process chains (multi-stations, i.e. spatiodimension) and machine state conditions over different parts in batches (i.e. temporal dimension), which makes quality prediction difficult. Current quality prediction methods have no consideration of equipment state degradation (ESD) or simply investigate a single machine. To improve the prediction accuracy of machining quality, a digital twin-driven part spatio-temporal quality prediction (DT-PSTQP) framework for multi-stage machining processes (MMP) is proposed with full considerations of multi-machine processes and multi-state machine degradation. The relationship graph analysis (RGA) is used to classify continuous ESD into limited discrete states to construct MMP reconstruction module. The DT-QPL module is a collection of quality prediction models that are trained with the refined sub-datasets obtained by MMP reconstruction. The proposed framework and the three models are validated through a thin-walled part production line. The results show that the proposed framework can help to improve the quality prediction average accuracy by 18.8% compared to the traditional framework without DT-PSTQP.

KEYWORDS:

• Machining quality prediction
• equipment state degradation
• digital twin
• deep adversarial learning

Acknowledgement

This work is financially supported in part by the National Natural Science Foundation of China (52275507), in part by the project supported by the National Natural Science Foundation of China (52075451).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the National Natural Science Foundation of China [52075451].