Engineering

Abstract: Vibration-based condition monitoring of beam-like metallic components requires vibration modes that are both sensitive to damage and consistent across different damage locations and severities. This study evaluates the damage sensitivity of the first eight bending-mode frequencies measured from 19 aluminum-alloy 6063 cantilever beams instrumented with magnetoelastic vibration sensors. The examined configurations comprised one undamaged reference beam and 18 damaged beams combining six crack-like defect locations with three defect depths. Normalized absolute frequency shifts were used to determine the mean sensitivity of each mode, its response to increasing defect depth, and the consistency of this response across the examined locations. The combined modal response increased progressively with defect depth, although the magnitude of the change depended on the selected mode and defect location. Mode f₇ exhibited the highest mean sensitivity of 1.234%and amonotonic-consistency ratio of C₇=1, indicating an increasing response with defect depth at all six locations. Modes f₂ and f₄ also demonstrated complete monotonic consistency but lower overall sensitivity, whereas f₅ showed comparatively high sensitivity with greater dependence on defect location. Mode f₈ presented the lowest sensitivity and consistency. These findings indicate that the available modes do not need to be treated equally. Prioritizing modes that combine high sensitivity with consistent depth-dependent behavior could simplify the processing and interpretation of modal information for preliminary condition monitoring of metallic industrial components. Further repeated measurements and validation under operational conditions are required before diagnostic thresholds can be established.

Abstract: Rear-discharge grain loss during rice harvesting is governed by interacting combine-harvester settings, yet studies often rely on fitted accuracy without uncertainty-aware validation. It was hypothesized that (H1) grain loss would exhibit significant nonlinear, quadratic, and interaction effects and (H2) validation would identify a stable operating region. A four-factor, five-level central composite block design comprising 30 runs evaluated the effect of forward speed, header-height set point, cleaning-fan speed, and feed-rate set point on grain loss. Response surface methodology - desirability function (RSM-DF) and artificial neural network - genetic algorithm (ANN-GA) yielded fitted minima of 23.58 and 14.70 kg/ha, respectively. Using a reconstructed 30-run analytical dataset, quadratic RSM, regularized quadratic regression, support vector regression (SVR), Gaussian process regression (GPR), and a weighted ensemble were compared through 20 repetitions of nested five-fold cross-validation. Permutation importance, SHapley Additive exPlanations (SHAP), and 1,000 residual-bootstrap models supported interpretation and robust optimization. Quadratic RSM generalized best (R² = 0.8751; RMSE = 6.18 kg ha⁻¹; MAE = 4.56 kg ha⁻¹), while cleaning-fan speed, header height, and forward speed dominated prediction. A robust plateau of 2.26 km/h, 150 mm, 1100 rpm, and 4.5–5.0 kg/s yielded 23.96 kg/ha (95% model-uncertainty interval: 21.70–26.41). The results indicate that both RSM-DF and ANN-GA can be utilized to model the processes of combine harvester operation to achieve better optimization results.

Abstract: Sucker rods operate under long-term cyclic loading while exposed to corrosive produced fluids, making corrosion–fatigue damage a major integrity concern. This study examined produced-fluid samples from 16 wells with sucker-rod failures and 16 corresponding failed rods from a Daqing oil production plant using ionic-composition analysis, SEM/EDS, bulk chemical analysis, metallography, axial tensile tests, exploratory fatigue tests, and field load analysis. The fluids were HCO₃⁻-dominated, with average HCO₃⁻, Cl⁻, and SO₄²⁻ concentrations of 345.7, 0.129, and 2.55 mg/L, respectively. Representative fracture surfaces showed corrosion pits, corrosion products, secondary cracks, and lamellar or step-like morphologies, while corroded microregions were enriched in O and contained detectable Cl and S. Local discontinuities, including decarburization, inclusions, and scratches, were also observed. Residual tensile strength varied substantially with service history, and exploratory fatigue results showed considerable scatter. The actual field fracture location did not coincide with the maximum nominal axial stress. The combined evidence supports a representative sequence of localized corrosion and pit formation, crack-susceptible region development, microcrack initiation, cyclic crack propagation, crack coalescence, loss of effective load-bearing area, and final rapid fracture.

Abstract: Background: An Anticipated Transient Without Scram (ATWS) represents one of the most severe conditions in nuclear reactor operations, escalating any concurrent malfunction into a Beyond Design Basis Accident (BDBA). When coupled with dual Large-Break Loss of Coolant Accidents (LBLOCA), station blackout, and initial fuel failure, the thermal-hydraulic stability and containment integrity of the reactor are severely compromised. Materials and Methods: In this study, the transient behavior and safety response of a Russian Gen-III+ VVER-1200 reactor were simulated using the IAEA-recommended Personal Computer Transient Analyzer (PCTran). An unmitigated accident scenario featuring simultaneous 1000cm² breaks in both the hot and cold legs, failure of the SCRAM mechanism, a total loss of AC power, and 5% fuel failure at power was evaluated over a 300-second simulation window. Model performance was benchmarked and validated against Final Safety Analysis Report (FSAR) core flow data. Results: The simulation demonstrates that the Departure from Nucleate Boiling Ratio (DNBR) dropped below 1.0 within 30 seconds, triggering an immediate boiling crisis due to sustained core power. Complete steam voiding across the reactor core occurred in 80 seconds. Consequently, reactor building pressure exceeded the 5-bar design threshold in only 90 seconds and reached 14.93 bar at 300 seconds, while containment temperatures peaked at 193.7 °C. Although peak fuel and cladding temperatures remained within design limits to prevent immediate core meltdown, the rapid containment over-pressurization severely restricts available operator response windows, posing a critical containment failure hazard. Conclusion: Simultaneous ATWS, dual-leg LBLOCA with fuel failure, and station blackout in a VVER-1200 induce a critical boiling crisis within 30 seconds and full core voiding in 80 seconds due to unmitigated core power. Although passive systems prevent immediate fuel and cladding meltdown, containment pressure breaches the 5-bar design threshold in just 90 seconds, highlighting that containment over-pressurization rather than core degradation is the most immediate threat requiring enhanced passive mitigation.

Abstract: Ground clutter introduces stationary or recurrent nonmeteorological echoes into weather-radar remote-sensing imagery and can be difficult to remove when archived products contain reflectivity alone. This study develops an image-level ground-clutter filter using Météo-France MeteoNet radar composites from northwestern France for 2016. Candidate screening and manual interpretation produced 18 clear-air events and 34 weather events, comprising 346,629 labelled pixels; ambiguous echoes were excluded. Ten reflectivity-derived intensity, spatial, temporal, and object features were evaluated with complete events held out. A random forest (RF) achieved clutter suppression of 0.763, weather-echo preservation of 0.952, and F1=0.833, outperforming reflectivity-only deterministic rules in their balance between removal and preservation. Component size and reflectivity were the leading permutation-importance variables, while temporal statistics provided a smaller complementary contribution. Rain-gauge data were used only for period screening, recurrent-clutter station confirmation, and an auxiliary downstream check. At nine confirmed stations, 4014 paired 30-min samples showed that RF gating reduced mean absolute error from 0.055 to 0.052 mm and false-alarm ratio from 0.277 to 0.181; root-mean-square error remained essentially unchanged (0.283 to 0.284 mm), while probability of detection increased from 0.635 to 0.738 and the critical success index increased from 0.511 to 0.635. These results show that the proposed reflectivity-only RF filter removes recurrent ground clutter while retaining most weather echoes, and the independent gauge comparison further shows that targeted filtering reduces false precipitation detections at confirmed clutter sites.

Abstract: Accurate measurement of ultra-fine dust concentration is crucial for effective dust control and safe mining. However, the precise detection of dust concentration during mining has not been fully addressed. This study established a controllable laboratory system to investigate the relative errors of two dust monitoring methods based on different measurement principles, and to analyze the relationship between particle attachment and measurement errors. Using a laser dust monitor and a double-head sampler, three concentration levels were tested on three representative dust samples (coal dust, carbon powder, and SiO₂). The relative errors of the two sampling methods were evaluated by the system, and the relationship between the actual values and the measured values was determined through fitting analysis. Additionally, the causes of particle attachment were further explored using characterization methods. The results showed that the relative error was larger at high concentrations. For the laser dust monitor, the relative error ranged from 85.30% to 97.39%, while for the double-head sampler, the relative error was between 46.18% and 82.05%. Particle size, dust characteristics, and attachment behavior significantly affect the accuracy of dust concentration measurement, providing a reference basis for subsequent mining operations and research.

Abstract: Natural fibers from different plants are widely used as reinforcing fillers in polymers because of ecological, economic and technical benefits with applications in all major industries. Since a typical drawback of incorporating natural fibers in polymers is the reduction of impact strength this work investigates ways to enhance impact strength. 30 wt. % of grass fibers from permanent pasture grasslands are compounded with polypropylene and varying amounts of polypropylene-graft-maleic anhydride with three different degrees of grafting as coupling agent and ethylene propylene diene rubber as impact modifier. The impact strength and the melt flow index are measured for each compound while the adhesion between fibers and matrix is investigated by scanning electron microscopy. It can be shown that around 5 wt. % of coupling agents are sufficient to increase impact strength to a saturation level while the impact modifier has no significant influence. By multiple extrusion cycles of a pure grass fiber polypropylene compound, it can be shown that the melt flow index increases constantly with the number of extrusion cycles while the impact strength remains constant after an initial increase indicating matrix degradation but enhanced fiber matrix coupling.

Abstract: The increasing integration of communication networks, intelligent electronic devices, distributed energy resources, and automated control has transformed the electric grid into a tightly coupled cyber-physical system. This transformation also creates opportunities for false-data injection attacks (FDIAs), in which adversaries manipulate measurements or telemetry to bias state estimation and influence operational decisions. Coordinated FDIAs are especially challenging because several measurements, devices, or communication paths can be manipulated in a spatially and temporally consistent manner. A detector that examines individual measurements in isolation may therefore fail to identify an attack that appears plausible at each local point while being harmful at system level. This paper develops a cyber-physical resilience framework for coordinated FDIAs. The framework extends protection beyond attack detection by organizing defenses into preparation, observation, detection, localization, response, recovery, and adaptation. It combines physics-based state estimation with data-driven anomaly detection, network and device telemetry, adaptive measurement trust, and operational safeguards. The paper also proposes a reproducible simulation protocol for evaluating the framework on benchmark power-system models under different attack coverage, duration, grid-condition, and communication-failure scenarios. Rather than reporting unperformed experiments, the manuscript defines the metrics, baselines, ablation studies, and reporting standards required for a credible empirical evaluation. Finally, it discusses how recent work on adversarially robust photovoltaic diagnosis can inform hybrid detector design while distinguishing photovoltaic fault diagnosis from grid-level FDIA detection. The result is a structured research agenda for designing smart-grid defenses that preserve safe operation, not merely classification accuracy.

Abstract: Under conditions of gradual depletion of gas-condensate field reserves and declining well production rates, one of the key challenges in operating offshore production facilities is improving the energy efficiency and reliability of process equipment, particularly compressor stations used for gas gathering, treatment, and transportation. Resource conservation and the adaptation of operating modes to changing production conditions require flexible compression systems capable of operating efficiently over a wide range of inlet flow rates and pressures. This article examines examples from three key regions where compressor stations have been adapted to declining production rates and reservoir pressures, and analyses the challenges encountered under actual operating conditions. A compressor model is proposed for the late stage of field development, during which the production rate of gas-condensate wells decreases from 5.0 million m³/day to 0.5 million m³/day over a ten-year period. The wellhead pressure and compressor suction pressure decline from 5.0 MPa to 0.1 MPa, while the discharge pressure remains at 5.5 MPa. Calculations were performed for the principal operating parameters, including gas production rate, wellhead and compressor suction pressures, the capacity of a single compressor under different configurations, the required number of compressors, and the total annual capacity of all operating compressors. The proposed compressor-unit reconfiguration scheme makes it possible to compress natural gas for at least nine years without requiring major upgrading or the construction of a new compressor facility. Unlike centrifugal compressors, which become inefficient when operating below 50% of their design capacity, the proposed configuration maintains effective operation at loads as low as 10% of the design capacity.

Abstract: The most promising class of optical gyroscope from the miniaturization point of view is the resonant optical gyroscope with a low-coherence radiation source. The present study briefly discusses the principles of operation of main types of such resonators. One of the most significant drawbacks of the resonant optical gyroscope with a low-coherence radiation source is its low energy efficiency. Only a small portion of the radiation carrying information about the angular velocity is directed from the source to the photodetector, specifically the radiation at frequencies corresponding to the eigenfrequencies of the ring resonator, the rest of the radiation being scattered on the unused ports of the resonator and converted into heat. It has been shown that the use of parallel-connected ring resonators can increase the energy efficiency of resonant gyroscopes with low-coherence radiation sources by an order of magnitude, from a few percent to tens of percent. It has also been theoretically demonstrated that this can reduce the contribution of shot noise and thermal noise of the photodiode by several times, and increase the sensitivity of the gyroscope.

Abstract: We consider the technical feasibility of a wirelessly distributed antenna array for C-UAS using directional microwave frequency energy. Importantly, the large possible size of the distributed antenna permits engagement of UAS and UAS swarms within the near-field region, enabling highly localized effects, ~ 5 meters at a range of 1 kilometer when operating at 1 GHz. In addition to near-field operation, combining N coherent emitters gives an N² enhancement of irradiance over that of a single emitter, enabling extension of system range by a factor N or a reduction in total system power requirement of N^-1 for to achieve an equivalent effect. Systems operating at 1-GHz require precise synchronization of order 100 picoseconds to ensure a high degree (> 90%) of coherence at target and therefore would require semiconductor power amplifiers rather than vacuum electron devices which are typically used as high power microwave sources. The authors develop a mathematical model implemented in software to explore the potential performance of a wirelessly distributed antenna for C-UAS and suggest some approaches for realization of such a technology.

Abstract: A safety margin derived from a vision–language verdict is only as valid as that verdict is recent, yet run-time safety filters model state noise, not the staleness of an intermittent label. We present AEGIS, a safe-action projection whose hold radius is an explicit function of a semantic belief and of the age of the verdict that produced it, with a conditional grasp-site separation result and a fresh-evidence feasibility property under stated assumptions. AEGIS runs on one fixed Intel RealSense D435i: localisation through a calibrated homography with a sub-millimetre mean fitting residual, a workspace-gated intruder detector fused with metric depth, and an on-device verdict returned in a fraction of a second. Beneath it a geometric layer checks and, when necessary, modifies commands against a set of geometric constraints, so it wraps any policy without retraining, reducing accumulated constraint cost in simulation without a significant loss of task success. In a buffer-matched comparison, indexing the margin by verdict age is significantly safer than trusting the latest verdict, raising the fraction of episodes free of any commanded approach to a person, and no constant belief reproduces the behaviour. On the physical Niryo NED3 Pro the logged radius expands as a verdict ages and collapses on a fresh one: AEGIS held positive measured clearance, one trust-latest run ended up inside the detected footprint, and the worst-case policy delivered least. Claims are reported against explicitly separated evidence tiers.

Abstract: Bangladesh faces three interlocking water quality crises: naturally occurring arsenic in shallow groundwater, salinity intrusion across the coastal southwest, and severe industrial and domestic pollution of the rivers around Dhaka. Conventional monitoring is too sparse, slow, and costly to manage these problems at national scale. This paper critically reviews how artificial intelligence (AI), and machine learning (ML) in particular, is being used to assess and improve water quality in the specific hydrological, climatic, and institutional context of Bangladesh, and asks what is still missing. We synthesise recent studies on arsenic risk mapping, coastal salinity forecasting, river water-quality-index (WQI) prediction, satellite and Internet-of-Things monitoring, and treatment optimisation, and summarise the governing equations of the dominant methods: tree ensembles, recurrent and physics-informed neural networks, and explainable AI. Reported performance commonly exceeds 0.9 (R²) for WQI regression and 0.9 accuracy for arsenic classification, yet most studies remain single-problem, single-region, and weakly validated across the monsoon cycle. We argue that the decisive gap is integration, and propose a national framework linking Bangladesh's monitoring agencies, remote sensing, and low-cost sensors to interpretable models and concrete management actions. We critically examine the principal barriers (data scarcity, seasonality, transferability, interpretability, and deployment capacity) and offer a prioritised research and policy roadmap aligned with Sustainable Development Goal 6. The contribution is a Bangladesh-specific synthesis and an actionable, interpretability-centred roadmap.

Abstract: Integrating precision agriculture (PA, a data-driven agricultural management system) with deep learning (DL) models can effectively support various activities, including yield prediction, crop health monitoring, field task automation, and decision-making. Taking advantage of such data-driven methodologies typically requires desktops, high-performance computing systems, and cloud clusters for data analysis, but their portability limits in-field applications. However, a single board computer, such as Raspberry Pi, offers a compact, lightweight, cost-efficient, easy-to-use, and feature-rich portable computing device which is ideal for in-field decision-making in PA applications. One such in-field application is crop growth stage classification for better crop management. Therefore, in this study, eight corn growth stages were classified using PhenoCam (near-surface [proximal] remote sensing network camera)imagery collected from ten PhenoCam sites. Four lightweight DL models were developed, ELiteCrop0, ELiteCrop1, ELiteCrop4, and MobNetCropV2, and evaluated across five image vertical clipping levels(0 %–40 %) using a supercomputer. The optimized model was subsequently deployed on a Raspberry Pi5 for edge inference. Model training accounted for the majority of the total CPU time, exceeding 97 %, while the testing times ranged from 0.01 min to 0.12 min, enabling real-time applications. Among the models, ELiteCrop0 achieved the most balanced performance with a confusion-matrix diagonal ratio(CMDR) of 0.93, followed by ELiteCrop1 (CMDR = 0.92). Overall, model performance decreased with increasing vertical clipping; therefore, a moderate image clipping (0 %–10 %) was recommended for improved computational efficiency. Analysis with a supercomputer produced an intrasite (same train sites)accuracy of 0.90–0.93 (Raspberry Pi: 0.78–0.81) and an intersite (new test sites) accuracy of 0.48–0.50(Raspberry Pi: 0.41–0.43), indicating challenges with model generalization. Raspberry Pi successfully processed ≈ 1000 images/min under safe operating conditions (68◦C). Future work should focus on extending the multi-site dataset to improve cross-site performance. Hence, this study presents a scalable and cost-effective solution for real-time corn growth stage monitoring in PA.

Abstract: Gait analysis is a robust, non-invasive method for evaluating motor function and detecting neurological dysfunctions, as slight abnormalities in walking patterns often indicate underlying medical conditions. With the increasing global prevalence of neurological and motor impairments, there is a rising need for scalable objective methods capable of supporting diagnosis, monitoring progression and guiding rehabilitation. In this work, we present a Spatial-Temporal Attention-Guided Network (STAG-Net) for pathological gait classification using the GAIT-IT dataset, including five gait types associated with different problems: scissor, spastic, steppage, propulsive, and normal. The dataset provides a challenging benchmark characterised by class imbalance, high inter-class similarity, and significant intra-subject variability. To address these challenges, we develop a hybrid architecture combining Convolutional Neural Networks for spatial feature extraction with Recurrent Neural Networks to model temporal gait dynamics. We propose a Bidirectional Long Short-Term Memory module that captures gait dynamics, enabling the model to better represent cyclical and symmetrical gait characteristics. Furthermore, we integrate a temporal attention mechanism to dynamically weight salient gait phases, supporting the network’s ability to discriminate between visually similar gait abnormalities. The resulting STAG-Net framework achieves 96.82% classification accuracy, demonstrating robustness in capturing gait differences and scalability for clinical and real-world applications.

Abstract: Deep saline aquifers serve as primary locations for storing large amounts of CO₂ but their injectivity faces a major threat due to salt accumulation in the reservoir pores. The research team conducts experiments to study how Brine chemistry affects salt precipitation, and they test different methods for removing these salts through laboratory experiments. The research team performed core-flooding tests on Bentheimer and Salt Wash North sandstone cores which they treated with mono- and divalent saturated (20 wt%) Brines (NaCl, KCl, CaCl₂, MgCl₂) before they injected supercritical CO₂ to create salt deposits and then tested low‑salinity Brine (0.5 wt%) and seawater (3.5 wt%) for their ability to remove these deposits. The research shows that salt deposits lead to lower porosity and permeability which causes restricted water movement and results in complete loss of injection ability. The process of salt buildup in divalent Brines generates more severe pore blocking than monovalent Brines which leads to faster CO₂ breakthrough. The research shows that low‑salinity Brine outperforms seawater for total permeability recovery because it produces recovery rates between 40.6 and 68.4 percent while seawater produces recovery rates between 7.4 and 17.2 percent under the same experimental conditions. The research establishes a direct connection between Brine chemical composition and salt formation which leads to decreased water flow through injection systems. The research methods followed identical patterns which demonstrated steady experimental results through multiple tests. The research findings received support from stable pressure values and flooding test repetition and SEM–EDX analysis which showed salt deposits forming in pores and disappearing during the treatment process. The research results demonstrate that salt buildup stands as the main factor which leads to reduced injection capacity. The study shows that low-salinity water serves as an effective treatment to restore permeability when compared to seawater.

Abstract: Accurate assessment of water status in woody crops is essential for optimizing irrigation management, particularly under Mediterranean conditions characterized by both high spatial and temporal variability. Traditional field-based measurements of stem water potential are reliable but present limitations for large-scale operational applications. In this study, a supervised machine learning approach based on Extreme Gradient Boosting was developed to estimate this metric in Mediterranean olive orchards by integrating high-resolution PlanetScope multispectral imagery with meteorological variables describing atmospheric evaporative demand. The model has been trained and evaluated using a dataset comprising 1,321 field-based stem water potential measurements collected during the 2021–2025 period, together with 129 predictive features, including PlanetScope-derived spectral bands and vegetation indices, as well as meteorological predictors. The results demonstrate a high level of accuracy, achieving a coefficient of determination of 0.84, a root mean square error of 0.27, and a mean absolute error of 0.21. Variable importance analysis indicates that air temperature, solar radiation, and reference evapotranspiration are the most influential predictors, while spectral information acts as a modulating factor incorporating effects related to canopy structure, vegetative vigor, and accumulated physiological responses. Furthermore, the combined use of spectral indices and base band reflectance values from PlanetScope, particularly those related to Photochemical Reflectance Index and Near Infrared, consistently improving model accuracy of 30% from previous proposed algorithms.

Abstract: Among low-lying sandy coasts, where erosion represents a major challenge, artificial beach nourishment is extensively used as mitigation strategy. This study investigates the long-term morphological evolution of Lido di Dante beach (Northern Emilia-Romagna, Italy), a highly subsiding and erosion-prone coastal sector characterized by different configurations of rigid defence structures. The analysis is based on 12 multi-temporal topo-bathymetric survey datasets acquired between 2007 and 2025, covering the complete monitoring sequences of three regional nourishment interventions (Progettone 2, 3, 4). An integrated methodological approach combining sediment budget calculations, shoreline evolution analysis, cross-shore profile assessment, and beach slope evaluation was applied to characterize the morphological response of the nourished system. Results show a consistent post-nourishment response among the three interventions, with effective beach recovery followed by sediment redistribution from the subaerial and intertidal zones towards the nearshore. However, the morphological evolution showed marked spatial variability, reflecting the influence of coastal defence structures on sediment redistribution pathways. The findings highlight the need for repeated nourishment interventions to maintain long-term beach stability at Lido di Dante. Furthermore, the study demonstrates the importance of accurate long-term topo-bathymetric monitoring for evaluating nourishment performance and supporting adaptive coastal management strategies in complex coastal environments.

Abstract: Nowadays, upper limb prosthesis acceptance remains low due to ineffective control techniques and the relative training methods. However, an engaging training method applied on early prosthetic fitting seems to have a positive impact on acceptance. To this aim, in this paper we present a Mixed Reality (MR) pre-fitting training system based on Microsoft HoloLens2, designed for users of the Hannes prosthetic hand. The system allows individuals with varying stump morphologies to control a holographic Hannes hand through a low-latency interface, replicating the physical architecture and motion of the real device in an immersive, portable environment. This pilot study presents the preliminary evaluation of the MR framework conducted with two limb difference participants (one naïve user and one experienced user in myoelectric prosthesis control) performing a novel bimanual Target Achievement Control (TAC) test. Their performance in controlling the real Hannes device was subsequently compared against a reference group of eight transradial limb difference individuals. Both MR-trained participants showed faster learning and better adaptation to the real prosthesis than the reference group. These preliminary findings suggest that the proposed MR framework, combined with a virtual bimanual TAC test, may improve learning and user experience, ultimately contributing to lower prosthesis abandonment rates.

Abstract: Collaborative robots have left the safety cage. In factories, operating theatres and care homes they now share physical and decision space with people. Physical safety is only one part of the ethical problem raised by that proximity. A second part concerns whether the person working beside the robot can still experience themselves as capable, recognised and in control of what they do. The question gains weight as robots become stronger and more precise, and as the AI systems directing them match or exceed human performance in narrow tasks and, plausibly, in broader ones. Technical standards such as ISO 10218:2025, which absorbs the former ISO/TS 15066, regulate force, speed and separation. The EU AI Act (Regulation 2024/1689) imposes human oversight obligations on high-risk systems. Neither instrument treats human dignity as a design property that can be specified, observed and audited, and neither addresses the risk that a more competent AI system diminishes human standing through cognitive rather than physical superiority. This paper proposes Dignity-by-Design (DbD), a framework that translates dignity-related concerns in robot ethics, drawing on Riek and Howard, on Sharkey and Sharkey, and on Nussbaum’s capabilities approach, into four operational principles: Informed Engagement, Override and Recovery, Affective Transparency, and Contextual Autonomy. Each principle carries auditable indicators, is anchored in Value Sensitive Design and is aligned with Article 14 of the AI Act. We work the framework through three deployment scenarios: an industrial cobot bench, an assistive robot for older adults, and an AI-supported professional decision environment. A research and certification roadmap follows, organised around a Dignity Impact Assessment. The contribution is conceptual rather than empirical. It offers a vocabulary and a measurement scaffold for moving dignity, as safety once moved, from ethical aspiration into engineering practice.