RT Journal Article
SR Electronic
A1 Kušnír, Ján
A1 Brindza, Jozef
A1 Demko, Michal
A1 Dominik, Filip
A1 Vrabeľ, Marek
T1 Design of Control System Architecture for Intelligent Fixture
JF Acta Mechanica Slovaca
YR 2025
VO 29
IS 3
SP 24
OP 29
DO 10.21496/ams.2025.025
UL https://www.actamechanica.sk/artkey/ams-202503-0003.php
AB We present a compact architecture for an intelligent fixture aimed at stabilizing the milling of thin-walled aerospace parts. The system fuses multi-sensor inputs tri-axial accelerometers (5-30 kHz) for vibration/chatter, strain or dynamometer signals (1-5 kHz) for cutting/clamping loads, and low-rate pressure/temperature (≤ 100 Hz) for thermal/fixturing state with an "edge to cloud" computing stack. A Raspberry Pi 5 performs synchronized windowing (0.5-1.0 s, 50% overlap), time-frequency analysis (STFT/wavelets), and lightweight features (RMS, crest factor, band energies, relative wavelet energy, entropy). Unsupervised detectors (one-class models, LSTM autoencoders) provide fast on-device deviation alerts, while server services handle training/retuning, dashboards, a model registry, and over-the-air deployment. Telemetry uses MQTT for efficient streaming and OPC UA for typed information models, PTP (IEEE-1588) aligns timestamps. A private QoS-aware 5G link carries features and event-driven raw snippets, supporting a split control strategy, safety-critical actions stay local, and supervisory updates (feeds/speeds, ae/ap, clamping) close over 5G. Anticipated benefits include improved accuracy and surface integrity, reduced scrap/rework, and better adaptability across parts and machines. Validation will proceed via stability-lobe experiments and trials on aerospace-grade components, with a planned upgrade to simultaneous-sampling IEPE acquisition and Acoustic Emission sensing for higher bandwidth and earlier wear detection.