Abstract
<title>Abstract</title> <p>Context: An aspect of smart manufacturing is ensuring operator safety while maintaining high productivity. In manufacturing, physiological stress and subjective safety perception can compromise performance. Objectives: This study used evolutionary computing techniques to optimise cutting parameters to minimize operator heart rate (HR) and perceived safety risk (Q_i), while maximizing material removal rate (MRR). Methods: Three metaheuristic optimisations methods were applied and compared: Genetic Algorithm (GA), Artificial Immune System (AIS), and Ant Colony Optimisation (ACO). The fitness function was implemented using Robust linear modelling with Huber’s T norm to model vibration, sound, HR, MRR and subjective safety perception across N=297 milling machining tasks. Optimal cutting parameters were identified for balancing physiological load, perceived risk, and productivity. Findings: Cutting axial depth and feed rate increase HR, vibration reduces physiological load, and HR is the primary determinant of perceived safety. Deterministic evaluation showed GA achieves the best safety-productivity trade-off, AIS provides a conservative yet feasible solution, and ACO exhibits instability. Across optimisation runs, AIS minimised HR (78 bpm) and safety risk (Q_i=1.01 ) while maintaining high MRR (460 mm^3⁄mm), GA prioritized safety at moderate productivity, and ACO maximised MRR (470 mm^3⁄mm) at higher HR (82 bpm). Feed rate dominated operator stress, whereas increased depth mitigated physiological load. Originality: This study integrates human-centric safety with productivity in milling operation using multi-objective metaheuristic optimisation, highlighting the combined influence of physiological and subjective metrics on machining performance and the efficacy of RLM for predictive modelling.</p>