Digital Twin (DTw) technology is a cornerstone of Industry 4.0, enabling real-time monitoring, predictive maintenance, and performance optimization across diverse industries. A key requirement for effective DTw implementation is high geometric fidelity—ensuring the digital model accurately represents the physical counterpart. Fidelity metrics provide a quantitative means to assess this alignment in terms of geometry, behavior, and performance. Among these, 3D shape descriptors play a central role in evaluating geometric fidelity, offering computational tools to measure shape similarity between physical and digital entities. This paper presents a comprehensive review of 3D shape descriptor methods and their applicability to geometric fidelity assessment in DTw systems. We introduce a structured taxonomy encompassing classical, structural, texture-based, and deep learning-based descriptors, and evaluate each in terms of transformation invariance, robustness to noise, computational efficiency, and suitability for various DTw applications. Building upon this analysis, we propose a conceptual fidelity metric that maps descriptor properties to the specific fidelity requirements of different application domains. This metric serves as a foundational framework for shape-based fidelity evaluation and supports the selection of appropriate descriptors based on system needs. Importantly, this work aligns with and contributes to the emerging ISO 30138 standardization initiative by offering a descriptor-driven approach to fidelity assessment. Through this integration of taxonomy, metric design, and standardization insight, this paper provides a roadmap for more consistent, scalable, and interoperable fidelity measurement in digital twin environments—particularly those demanding high precision and reliability.

This study introduces sPPO-CR, a policy-gradient deep reinforcement learning (DRL) algorithm for multi-objective (MO) and multi-condition multi-objective (MCMO) optimization. The method builds on a simplified single-step variant of Proximal Policy Optimization (PPO) and incorporates a Chebyshev-based reproduction mechanism with a log-likelihood rejection filter to safely integrate synthetic off-policy samples (sPPO-CR). sPPO-CR is benchmarked against sTD3-CR, a single-step variant of TD3, and three evolutionary baselines (NSGA-II, NSGA-III, and MOEA/D) across mathematical benchmarks and an airfoil shape optimization task. Results show that while evolutionary algorithms remain effective in standard MO settings, DRL methods, particularly sPPO-CR, offer improved performance in certain higher-dimensional and condition-dependent problems through effective generalization across both the design and condition spaces. Moreover, sPPO-CR achieves similar or superior performance to sTD3-CR while requiring an order of magnitude fewer network updates. Overall, the findings demonstrate that DRL-based optimizers provide a scalable and computationally efficient alternative for solving complex multi-condition multi-objective problems.

Wireless sensing inside rotating industrial machines is challenging due to signal attenuation, multipath propagation, and continuous sensor motion. Smart particles equipped with wireless communication and inertial sensors provide a promising approach for monitoring such systems. However, the reliability of wireless signal transmission under rotational dynamics remains insufficiently understood, and systematic approaches for sensor selection are lacking. This paper presents a link-budget-normalised experimental characterisation of three commercial smart particles, namely MetaMotionS (BLE), WitMotion BLE, and WitMotion Wi-Fi, in a bare 300 mm diameter by 310 mm deep metallic drum fitted with six triangular lifters, at rest and at 16, 18, and 20 RPM, corresponding to 20.7–25.9% of the critical speed and Froude numbers of 0.043–0.067. Because raw received power conflates transmit power with channel behaviour, the comparison is expressed as excess path loss above free space together with second-order fading statistics. Two particles of the same protocol class differ by 23.9 dB, of which at most 4 dB is attributable to the transmission of power across the documented range of both radios, establishing that device implementation rather than protocol class governs the ranking. Two particles logged simultaneously through a single receiver observe one channel realisation, and the correlation between their signal fluctuations is not significantly different from zero at any speed (r = +0.064, −0.098, −0.083), indicating device-specific rather than environmental fading. Under rotation, the Wi-Fi particle holds an RSSI standard deviation of 1.96 dB against 6.27 and 6.58 dB for the two BLE particles (Welch ANOVA, p < 0.001; Games–Howell post hoc, all pairwise comparisons p < 0.001; |Cliff’s δ| > 0.96). A bounded, dimensionless multi-criteria selection procedure over link margin, signal variability, cross-speed consistency, packet delivery, and energy per delivered packet is introduced; the ranking is invariant under weighted-sum and TOPSIS aggregation but inverts once endurance carries a weight above 0.35, which quantifies the trade-off between link quality and battery life. Coupling between the wireless and motion streams is examined by folding both onto rotation phase. A rotation-locked component in RSSI is detected in one of nine sensor–speed combinations, with a maximum modulation amplitude of 0.92 dB. The RSSI logging cadence of approximately 1 Hz resolves the drum fundamental but lies below the Nyquist requirement for the dominant motion band at 0.91–1.04 Hz, so joint wireless–motion studies of rotating machinery require RSSI logging at 5 Hz or above on a clock shared with the inertial unit.