Scheduling lag, path conflicts and resource overload are the common problems that a multi-robot collaborative assembly system faces under task dynamics changes, strong resource coupling and unexpected interruptions; scheduling mechanisms struggle to capture dynamic relations among entities and to verify strategy feasibility promptly. To tackle the problem, this paper proposes a dynamic scheduling optimization framework based on GraphSAGE (Graph Sample And aggreGatE) and digital twins. Through inductive learning, the framework embeds the dynamic relations of heterogeneous entities online and handles the real-time representation and scheduling of unseen nodes. A high-fidelity closed-loop simulation environment based on a digital twin verifies the feasibility and pre-evaluates the performance of each strategy before deployment, which prevents invalid or harmful scheduling instructions from reaching the physical system. Experiments evaluate the proposed approach under different task loads, robot counts and fault disturbance scenarios against five scheduling methods. The results indicate that for high task frequency, the average task response time is 2.6 s; the average task success rate is 95.0%; the number of path conflicts is less than 3; the rate of resource contention is less than 10%; and and the average rescheduling time under a 25% sudden failure rate is 0.85 s. Ablation experiments validate the graph embedding mechanism and the twin erification module separately. This study offers a secure and robust scheduling technique for multi-robot cooperation in high-speed dynamic intelligent manufacturing environments
Since there are various uncertain influences that affect the operation of the logistics supply chain, the existing supervision method cannot provide timely and efficient historical data, while the supply-demand relationship still needs enhancement. Therefore, this study puts forward a logistics supply chain risk dynamic monitoring and emergency collaborative decision-making framework based on Group Decision-Analytic Hierarchy Process and Multi-Agent Deep Reinforcement Learning. The framework aims to solve problems such as weak coordination in dynamic evolution and emergency decision-making of modern logistics supply chains, as well as insufficient coupling of multi-source risks. The framework first conducts risk level quantification and real-time monitoring. It combines Temporal Convolutional Network to extract spatiotemporal features and dynamically evaluate supply chain risks. It also introduces Exponentially Weighted Moving Average and other methods to generate emergency collaborative decision-making outputs. According to experimental results, the proposed model can monitor disruption events accurately with 94.02% and achieve a recall rate of 95.64%. In addition, the early warning delay is only 37.23 ms for monitoring the supply chain disruptions dataset under risk dynamic monitoring. In emergency coordination, the collaborative decision-making reward reaches 1.38 ± 0.10, the disruption loss reduction rate reaches 28.61% ± 8.11%, the resource utilization rate reaches 78.56%, and the task completion rate reaches 91.49%. The results clearly demonstrate that the model is superior to other models in terms of risk recognition accuracy, emergency response delay, resource use efficiency, and decision-making rewards for collaboration. The model is highly efficient in dynamic monitoring and provides effective emergency decision-making support for logistics chains. By improving risk perception, response coordination, and resource allocation efficiency, the proposed framework enhances operational resilience and disruption response capability of logistics chains.
This paper optimizes the milling operation on 3D-printed carbon fiber-reinforced acrylonitrile butadiene styrene (ABS+15% CF) composite using preference ranking organization method for enrichment of evaluations (PROMETHEE) in a multi-criteria decision making paradigm. Spindle speed (SS), feed rate (FR) and depth of cut (DOC) are treated as the milling parameters, while delamination factor, material removal rate, tool wear rate and surface roughness are the responses. Through main effects plots and analysis of variance results, influences of all the milling parameters on the responses are investigated. During optimization of the said milling process using PROMETHEE, effects of the six preference functions, i.e. usual, U-shape, V-shape, level, linear and Gaussian on the rankings of the conducted experiments are examined. It is revealed that V-shape, level, linear and Gaussian functions pick out trial number 25 (SS = 1000 rpm, FR = 600 mm/min and DOC = 0.2 mm) as the ideal choice, while, a parametric mix of SS = 1000 rpm, FR = 400 mm/min and DOC = 0.2 mm (experiment number 22) appears as the best trial for usual and U-shape functions. Considering the highest CRITIC weight allotted to MRR, experiment number 25 provides the optimal parametric combination for the considered milling operation on ABS+15% CF composite. Confirmation experiment at the optimal setting reproduces the observed responses and when evaluated by PROMETHEE against the other experiments, it retains the same rank as the initial trial for all six preference functions. The rankings of the alternative experiments are almost similar across the six preference functions of PROMETHEE. Thus, this data-driven approach can be effectively employed for parametric optimization of any of the manufacturing processes through simple calculation steps.
The present study aims to optimize the fabrication processing of Multi-Walled Carbon Nanotube (MWCNT)-filled glass fiber-reinforced polymer (GFRP) composites to enhance their tribological performance during the hand layup process. A structured experimental approach based on the Box-Behnken design within the Response Surface Methodology (RSM) framework was used to evaluate the influence of wt% MWCNT loading, applied load, and sliding velocity. The fabrication was performed, ensuring a uniform MWCNT dispersion via mechanical stirring and probe sonication. The performance of the composites was assessed in terms of their specific wear rate (SWR) and wear coefficients (WC). Statistical analysis, including ANOVA and Pareto charts, revealed that the load and sliding velocity significantly affected both the SWR and WC, whereas interactions such as the MWCNT content and sliding velocity also played a crucial role. Mathematical models were developed for SWR and WC and optimized using Elitist Teaching-Learning-Based Optimization (ETLBO). The optimal fabrication settings were found to be 1wt% MWCNT loading, 30 N load, and 2 m/s sliding velocity, yielding improved wear characteristics. Confirmatory tests validated the model predictions and showed minimal errors. Scanning electron microscopy further confirmed that optimal nanotube dispersion led to a more uniform and durable composite structure, supporting its potential for high-performance engineering applications. The proposed integrated approach demonstrates the effectiveness of combining RSM and ETLBO to develop high-performance MWCNT-GFRP composites with superior wear resistance. The findings of this study provide practical guidelines for fabricating durable, wear-resistant composite structures for automotive brake systems, aerospace joints, and industrial machinery parts, where reliability under continuous sliding conditions is essential.
Titanium alloy Ti–6Al–4V finds its suitability for aerospace applications along with compatibility with biological components. Creation of holes is a common requirement in the application of the alloy; these holes are created using abrasive water jet machining (AWJM) with variation of parameters against an L25 orthogonal array of Taguchi design. The work is aimed at obtaining a high-quality surface finish, a kerf taper angle as low as possible and a maximized machining rate (mm/s), relying on AWJM. The process parameters [water pressure, stand-off distance, abrasive flow rate and transverse speed (in terms of quality)] were also systematically evaluated to explore the combined effect on machining rate and surface integrity by plotting a contour plot of responses. Complex proportion assessment (COPRAS) multi-criteria decision-making technique that delivers balancing machining setup for number of process parameters which can provide maximum geometrical accuracy while maintaining satisfactory surface finish and production rate at 45 kPsi, abrasive flow rate at 300 g/min, and transverse speed in 75 mm/min at Quality 5 keeping stand-off distance as 2 mm. unique finding obtained by kerf taper measurement technique which indicate that the response is proportionally connected with depth of penetration, also not uniform throughout the depth. Experimental results exhibit the least deviation of machining rate and moderate deviation of surface roughness, while maximum variation is observed in the kerf taper angle and appears highly sensitive. Scanning electron microscopy (SEM) and digital microscopic morphology confirm the mechanism of surface integrity deviation and taper formation and validate the statistical findings.
