•  
  •  
 

Article Type

Original Study

Abstract

Achieving consistent surface integrity in CNC milling under varying materials and degraded tool conditions remains a critical challenge because fixed cutting parameters often induce regenerative chatter, accelerate tool wear, and cause surface and subsurface damage. This paper presents an integrated solution that combines an Ensemble Digital Twin (EDT), real-time in-process vibration monitoring, and a hybrid adaptive controller to maintain high surface quality while preserving productivity. The EDT fuses a physics-based milling dynamics model with a recurrent LSTM surrogate through a context-aware meta-learner that weights model outputs by process state (e.g., tool wear, engagement). A compact real-time monitoring pipeline computes a Chatter Severity Index (CSI) and summary spectral/time-domain features at 40 ms cadence. These inputs feed a Mamdani-type fuzzy supervisory controller whose defuzzified outputs are tracked by tuned PID actuators to adjust spindle speed and feed per tooth with sub-100 ms round-trip latency at the edge. Methodology includes numerical simulation, data-driven model training on a multi-source dataset, and on-machine experiments on representative alloys and tooling conditions. Results demonstrate that the ensemble yields substantially improved short-horizon predictions of chatter probability and surface roughness compared with single-model baselines, and that the integrated control system reduces vibration RMS and CSI, lowers mean Ra and Rz, and virtually eliminates deleterious white-layer formation while maintaining a material removal rate close to the baseline. The proposed framework offers a practical pathway for deploying predictive, closed-loop quality control in smart machining environments, reducing scrap and rework and enabling safer, higher-yield production under challenging operating conditions.

Keywords

CNC milling, Ensemble digital twin, Real-time adaptive control, Chatter detection, Surface integrity

Download

Share

COinS