Advanced Engineering Informatics最新文献

{"title":"Empowering LLMs by hybrid retrieval-augmented generation for domain-centric Q&A in smart manufacturing","authors":"Yuwei Wan ,&nbsp;Zheyuan Chen ,&nbsp;Ying Liu ,&nbsp;Chong Chen ,&nbsp;Michael Packianather","doi":"10.1016/j.aei.2025.103212","DOIUrl":"10.1016/j.aei.2025.103212","url":null,"abstract":"<div><div>Large language models (LLMs) have shown remarkable performances in generic question-answering (QA) but often suffer from domain gaps and outdated knowledge in smart manufacturing (SM). Retrieval-augmented generation (RAG) based on LLMs has emerged as a potential approach by incorporating an external knowledge base. However, conventional vector-based RAG delivers rapid responses but often returns contextually vague results, while knowledge graph (KG)-based methods offer structured relational reasoning at the expense of scalability and efficiency. To address these challenges, a hybrid KG-Vector RAG framework that systematically integrates structured KG metadata with unstructured vector retrieval is proposed. Firstly, a metadata-enriched KG was constructed from domain corpora by systematically extracting and indexing structured information to capture essential domain-specific relationships. Secondly, semantic alignment was achieved by injecting domain-specific constraints to refine and enhance the contextual relevance of the knowledge representations. Lastly, a layered hybrid retrieval strategy was employed that combined the explicit reasoning capabilities of the KG with the efficient search power of vector-based similarity methods, and the resulting outputs were integrated via prompt engineering to generate comprehensive, context-aware responses. Evaluated on design for additive manufacturing (DfAM) tasks, the proposed approach achieved 77.8% exact match accuracy and 76.5% context precision. This study establishes a new paradigm for industrial LLM systems, which demonstrates that hybrid symbolic-neural architectures can overcome the precision-scalability trade-off in mission-critical manufacturing applications. Experimental results indicated that integrating structured KG information with vector-based retrieval and prompt engineering can enhance retrieval accuracy, contextual relevance, and efficiency in LLM-based Q&amp;A systems for SM.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103212"},"PeriodicalIF":8.0,"publicationDate":"2025-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143463982","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A novel reinforced incomplete cyber-physics ensemble with error compensation learning for within-batch quality prediction","authors":"Yi Shan Lee ,&nbsp;Junghui Chen","doi":"10.1016/j.aei.2025.103172","DOIUrl":"10.1016/j.aei.2025.103172","url":null,"abstract":"<div><div>This study addresses the challenge of real-time quality monitoring in batch operation by emphasizing the significance of within-batch quality estimation. While data-driven machine learning models are easy to construct, they often lack reliability and interpretability when dealing with sparse quality data. Conversely, first-principles models (FPMs) are interpretable but struggle with accuracy and adaptability to changing conditions. To overcome these issues, a three-phase reinforced incomplete cyber-physical ensemble plus error compensation learning (RICPE-P-ECL) method is proposed. This method enhances the adaptability of the incomplete cyber-physical model (IncompCPM), which relies on partially-available FPMs, for online quality prediction under varying conditions. The innovation in RICPE-P-ECL lies in its ensemble design and error compensation strategy. Phase 1 constructs IncompCPMs to predict quality for each operating condition, creating base models for ensemble learning. Phase 2 combines these IncompCPMs, with real-time information assigning weights to each model. Phase 3 involves an error compensation agent that adjusts the real-time ensemble prediction, addressing the limitations of FPMs and sparse data. The method is evaluated using a fed-batch bioreactor as the process model, and the results demonstrate that RICPE-P-ECL outperforms traditional data-driven models such as semi-supervised latent dynamic variational autoencoder and semi supervised dual attentioned latent dynamic complementary state space model, achieving R² values close to 1 for real-time within-batch quality prediction across five new testing conditions.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103172"},"PeriodicalIF":8.0,"publicationDate":"2025-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143463984","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Inverse design of lattice structures with target mechanical performance via generative adversarial networks considering the effect of process parameters","authors":"Chenglong Duan,&nbsp;Dazhong Wu","doi":"10.1016/j.aei.2025.103221","DOIUrl":"10.1016/j.aei.2025.103221","url":null,"abstract":"<div><div>While generative artificial intelligence has been used to design materials and structures for additive manufacturing, current techniques can only generate design parameters. However, not only design parameters but also additive manufacturing (AM) process parameters affect the mechanical properties of additively manufactured materials. To address this issue, we introduce an auxiliary classifier generative adversarial network (ACGAN)-based computational framework that generates both design and AM process parameters to fabricate lattice structures with target mechanical performance. The computational framework consists of two ACGAN models, including a generative model called InverseACGAN and a forward predictive model called ForwardACGAN. The generative model generates critical design parameters of the lattice structures, including line distance, layer height, and infill pattern, as well as AM process parameters, including print speed and print temperature, based on target mechanical properties (i.e., porosity and compressive modulus). The forward predictive model predicts the mechanical properties of the lattice structures generated by the generative model. The experimental results show that the porosity and compressive modulus of the lattice structures designed by ACGAN are in good agreement with the target porosity and compressive modulus. The average mean absolute percentage errors between target and actual porosity, and target and actual compressive modulus are 6.481% and 10.208%, respectively.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103221"},"PeriodicalIF":8.0,"publicationDate":"2025-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143463341","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An exact algorithm for RAP with k-out-of-n subsystems and heterogeneous components under mixed and K-mixed redundancy strategies","authors":"Jiangang Li ,&nbsp;Dan Wang ,&nbsp;Haoxiang Yang ,&nbsp;Mingli Liu ,&nbsp;Shubin Si","doi":"10.1016/j.aei.2025.103163","DOIUrl":"10.1016/j.aei.2025.103163","url":null,"abstract":"<div><div>Redundancy design is a widely used technique for enhancing system reliability across various industries, including aerospace and manufacturing. Consequently, the redundancy allocation problem (RAP) has attracted considerable attention in the field of reliability engineering. The RAP seeks to determine an optimal redundancy scheme for each subsystem under resource constraints to maximize system reliability. However, existing RAP models and exact algorithms are predominantly confined to simple 1-out-of-<em>n</em> subsystems or single optimization strategies, thereby limiting the optimization potential and failing to adequately address the engineering requirements. This paper introduces a model and an exact algorithm for RAP with <em>k</em>-out-of-<em>n</em> subsystems and heterogeneous components under mixed and K-mixed redundancy strategies. The model employs a continuous time Markov chain method to calculate subsystem reliability exactly. A dynamic programming (DP) algorithm based on super component and sparse node strategies is designed to obtain the exact solution for RAP. Numerical experiments confirm that all benchmark test problems reported in the literature are exactly solved by the proposed DP. The experiment results demonstrate that the proposed RAP model offers high flexibility and potential for reliability optimization. Additionally, owing to the generality of the problem considered, the proposed DP also exactly solves other RAP models with 1-out-of-<em>n</em> subsystems and simplified redundancy strategies, which provides a more generalized framework for redundancy optimization. Finally, the research’s applicability in reliability engineering is validated through an optimization case study of a natural gas compressor pipeline system.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103163"},"PeriodicalIF":8.0,"publicationDate":"2025-02-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143445305","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Research on multi-step ahead prediction method for tool wear based on MSTCN-SBiGRU-MHA","authors":"Jing Xue,&nbsp;Yaonan Cheng,&nbsp;Wenjie Zhai,&nbsp;Xingwei Zhou,&nbsp;Shilong Zhou","doi":"10.1016/j.aei.2025.103219","DOIUrl":"10.1016/j.aei.2025.103219","url":null,"abstract":"<div><div>Tool wear monitoring (TWM), as an important component of modern intelligent processing, faces significant challenges related to accuracy and long-term predictability. This research proposes a method for the precise and reliable multi-step prediction of tool wear. First, a dual-indicator feature screening scheme is proposed. The constructed sensitive features can describe the tool wear condition from multiple perspectives. Further, the MSTCN-SBiGRU-MHA model is developed to effectively analyze time series data by incorporating three key modules. The synergistic interaction among these three modules contributes to the model’s superior performance in complex time series prediction tasks. Finally, the multi-step prediction approach is integrated with interval prediction, and the validity of the resultant predictions is substantiated through milling experiments. Ablation experiments and the SHAP method are used to analyze the contribution of different modules and features to the model’s performance. Comparative experiments show that the model’s R2 for predicting the next 1, 5, and 10 steps across various datasets exceeded 0.86, significantly outperforming the SLSTM, SGRU, SBiLSTM-AT, and CNN-LSTM models. Accurate advance prediction of tool wear is crucial for developing an intelligent early warning system, ensuring high-quality production, reducing operational and maintenance costs, and enhancing machining safety.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103219"},"PeriodicalIF":8.0,"publicationDate":"2025-02-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143454465","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Knowledge transfer from simple to complex: A safe and efficient reinforcement learning framework for autonomous driving decision-making","authors":"Rongliang Zhou ,&nbsp;Jiakun Huang ,&nbsp;Mingjun Li ,&nbsp;Hepeng Li ,&nbsp;Haotian Cao ,&nbsp;Xiaolin Song","doi":"10.1016/j.aei.2025.103188","DOIUrl":"10.1016/j.aei.2025.103188","url":null,"abstract":"<div><div>A safe and efficient decision-making system is crucial for autonomous vehicles. However, the complexity of driving environments often limits the effectiveness of many rule-based and machine learning approaches. Reinforcement learning (RL), with its robust self-learning capabilities and adaptability to diverse environments, offers a promising solution. Despite this, concerns about safety and efficiency during the training phase have hindered its widespread adoption. To address these challenges, we propose a novel RL framework, Simple to Complex Collaborative Decision (S2CD), based on the Teacher–Student Framework (TSF) to facilitate safe and efficient knowledge transfer. In this approach, the teacher model is first trained rapidly in a lightweight simulation environment. During the training of the student model in more complex environments, the teacher evaluates the student’s selected actions to prevent suboptimal behavior. Besides, to enhance performance further, we introduce an RL algorithm called Adaptive Clipping Proximal Policy Optimization Plus (ACPPO+), which combines samples from both teacher and student policies while utilizing dynamic clipping strategies based on sample importance. This approach improves sample efficiency and mitigates data imbalance. Additionally, Kullback–Leibler (KL) divergence is employed as a policy constraint to accelerate the student’s learning process. A gradual weaning strategy is then used to enable the student to explore independently, overcoming the limitations of the teacher. Moreover, to provide model interpretability, the Layer-wise Relevance Propagation (LRP) technique is applied. Simulation experiments conducted in highway lane-change scenarios demonstrate that S2CD significantly enhances training efficiency and safety while reducing training costs. Even when guided by suboptimal teachers, the student consistently outperforms expectations, showcasing the robustness and effectiveness of the S2CD framework.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103188"},"PeriodicalIF":8.0,"publicationDate":"2025-02-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143454467","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Constraint programming-based layered method for integrated process planning and scheduling in extensive flexible manufacturing","authors":"Mengya Zhang,&nbsp;Xinyu Li,&nbsp;Liang Gao,&nbsp;Qihao Liu","doi":"10.1016/j.aei.2025.103210","DOIUrl":"10.1016/j.aei.2025.103210","url":null,"abstract":"<div><div>Current extensive flexible manufacturing systems are characterized by high flexibility and large problem sizes, which present significant challenges to manufacturing efficiency. The integrated process planning and scheduling (IPPS) is a significant issue in this context. Due to the complexity of the integration problem, various approximation algorithms have been developed to tackle it, though this often demands considerable designer expertise and parameter tuning. This paper proposes a constraint programming (CP)-based method that can solve the large-scale IPPS problem in extensive flexible manufacturing. Firstly, this paper proposes a CP model which enriches the variable decision-making for flexible processes. Based on this, this paper presents a hybrid layered constraint programming (HLCP) method, which decomposes the complete CP model into multiple models of sub-problems and solves these models iteratively to reduce the solution difficulty. It contains multiple sets of model relaxation and repair stages. Experiments on benchmark instances confirm that the proposed method reaches all optimal solutions, and surpasses previous results on 9 instances. Next, the proposed methods are tested on 35 sets of large-scale instances with up to 8000 operations, and the results show that the minimum gap can be obtained compared to the existing methods. Moreover, the proposed HLCP method is able to reduce the gap by an average of 9.03% within a reasonable time compared to the single-model approach.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103210"},"PeriodicalIF":8.0,"publicationDate":"2025-02-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143454464","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A hybrid two-way fluid-solid interaction method for intermittent fluid domains: A case study on peristaltic pumps","authors":"Qingye Li ,&nbsp;Xinxin Li ,&nbsp;Yuxue Li ,&nbsp;Xueguan Song ,&nbsp;De Li ,&nbsp;Yanfeng Zhang ,&nbsp;Yan Peng","doi":"10.1016/j.aei.2025.103191","DOIUrl":"10.1016/j.aei.2025.103191","url":null,"abstract":"<div><div>In this paper, a hybrid two-way fluid–solid interaction method (HTFSIM) is proposed to overcome the limitations of conventional two-way fluid–solid interaction method (CTFSIM) in simulating intermittent fluid domains, providing a more detailed understanding of the flow pulsation mechanism of peristaltic pumps. The HTFSIM distinguishes between intermittent and continuous fluid domains based on the peristaltic pump’s operating principle. By combining point cloud 3D reconstruction of hyper-elastic structures from finite element calculations with traditional two-way fluid–solid coupling, the flow in these domains is calculated separately and then superimposed to capture the flow fluctuations of the peristaltic pump cycle. Comparison of computational and experimental results with the CTFSIM demonstrates that the HTFSIM achieves higher computational accuracy and efficiency. Furthermore, the results regarding the contribution of individual rollers to the flow rate indicate that the flow rate variation caused by Roller 2 follows an asymmetric sinusoidal distribution, which influences the upper limit of the peristaltic pump outlet flow rate. Meanwhile, the reflux induced by Roller 1 affects the lower limit of the outlet flow rate. These findings are crucial for understanding the mechanism behind the flow pulsations in peristaltic pumps.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103191"},"PeriodicalIF":8.0,"publicationDate":"2025-02-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143445304","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A multistage stochastic programming approach for drone-supported last-mile humanitarian logistics system planning","authors":"Zhongyi Jin ,&nbsp;Kam K.H. Ng ,&nbsp;Chenliang Zhang ,&nbsp;Y.Y. Chan ,&nbsp;Yichen Qin","doi":"10.1016/j.aei.2025.103201","DOIUrl":"10.1016/j.aei.2025.103201","url":null,"abstract":"<div><div>Drone-supported last-mile humanitarian logistics systems play a crucial role in efficiently delivering essential relief items during disasters. In contrast to conventional truck-based transportation methods, drones provide a versatile and rapid transportation alternative. They are capable of navigating challenging terrain and bypassing damaged infrastructure. However, establishing an effective drone-supported last-mile humanitarian logistics system faces various challenges. This study introduces a novel approach to address these challenges by proposing a drone-supported last-mile humanitarian logistics system planning (DLHLSP) problem. The DLHLSP problem involves decision-making for both pre-disaster and post-disaster phases, taking into account the unique characteristics of drone-based delivery operations and uncertain demands. In the pre-disaster phase, decisions include determining drone-supported relief facility locations, drone deployment strategies, and drone visit schedules to disaster sites. Post-disaster decisions focus on inventory management, relief item procurement, and drone-based delivery operations. To capture the demand uncertainty in chaotic disaster environment, we establish a multistage stochastic programming model incorporating nonanticipativity constraints to make decisions at each stage without knowledge of the demand information in future time periods. Next, we employ the Benders decomposition algorithm to obtain exact solutions. Furthermore, we perform numerical experiments to verify the exact algorithm using randomly generated numerical instances. The results show that the algorithm significantly outperforms the Gurobi solver and could solve the problem of practical scale. Finally, the study validates the proposed model based on a case study of the Lushan earthquake in China and provides several managerial implications and insights. Overall, this research contributes to the field of humanitarian logistics by offering a comprehensive framework for the planning of drone-supported last-mile humanitarian logistics systems.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103201"},"PeriodicalIF":8.0,"publicationDate":"2025-02-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143454466","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"ADCL: An attention feature enhancement network based on adversarial contrastive learning for short text classification","authors":"Shun Su ,&nbsp;Dangguo Shao ,&nbsp;Lei Ma ,&nbsp;Sanli Yi ,&nbsp;Ziwei Yang","doi":"10.1016/j.aei.2025.103202","DOIUrl":"10.1016/j.aei.2025.103202","url":null,"abstract":"<div><div>Supervised Contrastive Learning (SCL) has emerged as a powerful approach for improving model performance in text classification tasks, particularly in few-shot learning scenarios. However, existing SCL methods predominantly focus on the contrastive relationships between positive and negative samples, often neglecting the intrinsic semantic features of individual samples. This limitation can introduce training biases, especially when labeled data are scarce. Additionally, the intrinsic feature sparsity of short texts further aggravates this issue, hindering the extraction of discriminative and robust representations. To address these challenges, we propose a Label-aware Attention-based Adversarial Contrastive Learning Network (ADCL). The model incorporates a bidirectional contrastive learning framework that leverages cross-attention layers to enhance interactions between label and document representations. Moreover, adversarial learning is employed to optimize the backpropagation of contrastive learning gradients, effectively decoupling sample embeddings from label-specific features. Compared to prior methods, ADCL not only emphasizes contrasts between positive and negative samples but also prioritizes the intrinsic semantic information of individual samples during the learning process. We conduct comprehensive experiments from both full-shot and few-shot learning perspectives on five benchmark short-text datasets: SST-2, SUBJ, TREC, PC, and CR. The results demonstrate that ADCL consistently outperforms existing contrastive learning methods, achieving superior average accuracy across the majority of tasks.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103202"},"PeriodicalIF":8.0,"publicationDate":"2025-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143445303","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}