A hybrid information fusion method for failure tracing of CNC machine tools based on meta-action failure-parameter-cause-target analysis

Accurate failure tracing is essential for closed-loop control of CNC machine tools. To address poor interpretability, strong subjectivity, and complex modeling in existing methods, a meta-action failure tracing method is proposed. Meta-actions are obtained through “Function-Motion-Action” decomposition. Dynamic models are established for four motion performance parameters of the meta-action: motion trajectory, speed, position, and speed fluctuation coefficient. A mapping between meta-action failure modes and basic parameters is constructed. An analysis of the complete failure causes and targets is conducted, and an interpretable Failure-Parameter-Cause-Target (FPCT) framework is developed. To handle limited data and multi-source information, the Best-Worst Method is used to extract expert evidence, which is further objectively corrected using Sobol-based importance coefficients. Dempster-Shafer theory is applied to fuse multi-source evidence and identify the most likely cause and the corresponding target. A case study of a CNC pallet changer shows that ''Fluctuation excessive'' and ''Displacement insufficient'' failures of the bracket meta-action are caused by power path failure and gear shaft and rack loosening, respectively. Among 18 cases, 16 reasoning results match the actual causes, and the remaining 2 rank within the top 3. Sensitivity analysis demonstrates good stability. The proposed method provides reliable support for failure tracing and maintenance decision-making.