With growing concerns about resource scarcity and environmental degradation, remanufacturing of end-of-life (EoL) products within the circular economy is attracting increasing attention. Remanufacturing can preserve most of the original manufacturing value and materials while transforming EoL products into like-new condition. However, the variability and uncertainty of EoL products make remanufacturing highly dependent on human expertise. Although recent advances in robotics and artificial intelligence have improved automation in isolated remanufacturing tasks, existing methods are often task-specific and struggle to generalize to heterogeneous EoL conditions. Recently, large language models (LLMs) have demonstrated remarkable capabilities in learning from massive, unstructured datasets, generating expert-level output across various tasks, and communicating with humans in natural language for interpretation. These advantages can align closely with the complex demands of remanufacturing, thereby mitigating the reliance on specialized expertise. However, their roles and research progress in this domain remain underexplored. In this paper, we present a forward-looking review and analysis of the role of LLMs in remanufacturing automation, grounded in a brief critical review of existing LLM-related studies relevant to remanufacturing. Building on this foundation, we introduce ReManGPT as a conceptual framework and use three representative case studies to illustrate selected modules of the framework in practical remanufacturing scenarios. We also analyze three representative remanufacturing applications, electric vehicle batteries, electronic waste, and electric motors, to illustrate how the proposed framework could address their domain-specific challenges. Finally, we discuss the current barriers to deploying this framework in practice and outline future research directions, including LLM-assisted human operation and language–action models for robotic automation.

Remanufacturing is fundamentally more challenging than traditional manufacturing due to the significant uncertainty, variability, and incompleteness inherent in end-of-life (EoL) products. At the same time, it has become increasingly essential and urgent for facilitating a circular economy, driven by the growing volume of discarded electronic products and the escalating scarcity of critical materials. In this paper, we review the existing literature and examine the key challenges as well as emerging opportunities in intelligent automation for EoL electronics remanufacturing, providing a comprehensive overview of how robotics, control, and artificial intelligence (AI) can jointly enable scalable, safe, and intelligent remanufacturing systems. This paper starts with the definition, scope, and motivation of remanufacturing within the context of a circular economy, highlighting its societal and environmental significance. Then it delves into intelligent automation approaches for disassembly, inspection, sorting, and component reprocessing in this domain, covering advanced methods for multimodal perception, decision-making under uncertainty, flexible planning algorithms, and force-aware manipulation. The paper further reviews several emerging techniques, including large foundation models, human-in-the-loop integration, and digital twins that have the potential to support future research in this area. By integrating these topics, we aim to illustrate how next-generation remanufacturing systems can achieve robust, adaptable, and efficient operation in the face of complex real-world challenges.

Precise segmentation of irregular and densely arranged components is essential for robotic disassembly and material recovery in electronic waste (e-waste) recycling. This study evaluates the impact of model architecture and scale on segmentation performance by comparing SAM2, a transformer-based vision model, with the lightweight YOLOv8 network. Both models were trained and tested on a newly collected dataset of 1,456 annotated RGB images of laptop components including logic boards, heat sinks, and fans, captured under varying illumination and orientation conditions. Data augmentation techniques, such as random rotation, flipping, and cropping, were applied to improve model robustness. YOLOv8 achieved higher segmentation accuracy (mAP50 = 98.8

The rapid growth of electric vehicles has created an urgent need for practical end-of-life lithium-ion battery recycling systems that are technically competent, economically viable, and environmentally sustainable. To clarify these growing technical and systemic challenges, a detailed synthesis of the existing body of knowledge is required. This review systematically analyzes the EV battery recycling literature retrieved from Scopus, Web of Science, and Google Scholar. From an initial set of about 1,700 publications, 130 studies were selected through structured screening for methodological relevance and technical analyses. The review is structured around four major topics: (1) collection and infrastructure planning, (2) recycling technologies, (3) digital and automation technologies, and (4) lifecycle and techno-economic assessments. For collection and infrastructure planning, the literature on game-theoretic and reverse logistics models show that coordinated networks and policy-driven incentives considerably improve collection outcomes and economic performance. In the area of recycling technologies, studies discuss that mechanical disassembly is mainly manual due to design heterogeneity and safety risks; while emerging robotic solutions deliver operational improvements. Pyrometallurgy provides robust throughput but is energy-intensive, hydrometallurgy facilitates selective recovery, supercritical and electrochemical routes offer cleaner options, and direct regeneration supports closed-loop circularity but requires scale-up. About digital and automation technologies, AI supports advanced diagnostics, robotics increases flexibility, digital twins facilitate predictive control, and digital product passports advance traceability but face governance and standardization challenges. Also, in the lifecycle assessment and techno-economic assessments domain, studies suggest that logistics and collection rates dominate cost and emission profiles, with siting optimization and automation driving measurable improvements. The synthesis of literature also identifies three future research directions: intelligent hybrid systems, information recovery systems, and resilient value networks which emphasize the need for digitally connected policy-aligned recycling infrastructures.

As our reliance on electronic devices grows, electronic waste (e-waste) has become the fastest-growing waste stream, posing significant threats to the environment and public health. Aligned with Industry 5.0, e-waste disassembly promotes sustainability through recycling valuable materials. However, the process is predominantly performed by humans with substantial occupational hazards. To address this, adopting collaborative robots (cobots) to complement humans has received attention. This study proposes a three-step framework for implementing human-robot collaboration in e-waste disassembly: task modelling, task allocation and scheduling, and ergonomics evaluation. A case study on desktop computer disassembly was conducted. Hierarchical task analysis (HTA) was used for task modelling. An optimisation-based method considering disassembly cost, operator safety, and task complexity was adopted for task allocation and scheduling. Ergonomics evaluation with 22 participants revealed that performing the task with a cobot significantly reduced workload and ergonomic risks for humans compared to disassembling alone.

Automating disassembly of critical components from end-of-life (EoL) desktops, such as high-value items like RAM modules and CPUs, as well as sensitive parts like hard disk drives, remains challenging due to the inherent variability and uncertainty of these products. Moreover, their disassembly requires sequential, precise, and dexterous operations, further increasing the complexity of automation. Current robotic disassembly processes are typically divided into several stages: perception, sequence planning, task planning, motion planning, and manipulation. Each stage requires explicit modeling, which limits generalization to unfamiliar scenarios. Recent development of vision-language-action (VLA) models has presented an end-to-end approach for general robotic manipulation tasks. Although VLAs have demonstrated promising performance on simple tasks, the feasibility of applying such models to complex disassembly remains largely unexplored. In this paper, we collected a customized dataset for robotic RAM and CPU disassembly and used it to fine-tune two well-established VLA approaches, OpenVLA and OpenVLA-OFT, as a case study. We divided the whole disassembly task into several small steps, and our preliminary experimental results indicate that the fine-tuned VLA models can faithfully complete multiple early steps but struggle with certain critical subtasks, leading to task failure. However, we observed that a simple hybrid strategy that combines VLA with a rule-based controller can successfully perform the entire disassembly operation. These findings highlight the current limitations of VLA models in handling the dexterity and precision required for robotic EoL product disassembly. By offering a detailed analysis of the observed results, this study provides insights that may inform future research to address current challenges and advance end-to-end robotic automated disassembly.

The global acceleration of electronic waste (e-waste) generation has created significant environmental, economic, and social challenges. Emerging technologies and shorter product lifespans are expected to intensify this growth. Despite the potential for material recovery, only a small fraction of e-waste is formally recycled, with a significant loss of critical resources such as rare earth elements and increased environmental degradation. Although prior studies address specific economic or environmental dimensions of e-waste management, detailed evaluations of recycling technologies from all three sustainability pillars are limited. This paper uses a structured sustainability framework to review five major recycling processes, including physical disassembly, pyrolysis, hydrometallurgy, biometallurgical treatment, and supercritical fluid technology. Social implications include occupational health and safety risks, public health impacts, and socioeconomic disruptions associated with transitions from an informal to a formal system. The results show physical disassembly and hydrometallurgical methods are widely used, however, they create considerable health risks and require better environmental impact data. Biometallurgical approaches have lower environmental toxicity but are constrained by limited scalability and process efficiency. Pyrolysis provides partial energy recovery but generates concerns over pollutant emissions and worker safety. Supercritical fluid technologies have high technical promise, however, their economic and operational viability are underdeveloped. The paper proposes a roadmap for advancing e-waste recycling systems by identifying data gaps and technology-specific opportunities for sustainable scale-up.

This study provides an AI-based product durability assessment framework for estimating the longevity of products based on historical repair logs. It uses a repair and maintenance dataset of medical equipment to develop a scoring model that assesses product longevity and failure trends despite limited data attributes. The dataset includes work order numbers, asset identifiers, equipment descriptions, manufacturers, models, serial numbers, service dates, and repair determinations. The study applies machine learning techniques to analyze patterns in failure frequency, time between failures, equipment types, and manufacturer-specific issues to develop a durability score that aids in optimizing maintenance scheduling and resource allocation. To improve the model, feature engineering is used on categorical fields, such as equipment type, manufacturer, and model, to change these into predictive features that demonstrate failure likelihood and product lifespan trends. Moreover, temporal analysis on repair dates shows the long-term reliability of specific models over time and further improves the durability score. The study correlates repair outcomes (determination descriptions) with failure frequency and time to failure to identify high-risk equipment and provide recommendations for preventive maintenance. The analyses demonstrate that, despite limited data attributes, it is possible to generate practical information about product longevity using AI models trained on categorical and temporal data.

Battery lifetime and reliability depend on accurate state-of-health (SOH) estimation, while complex degradation mechanisms and varying operating conditions strengthen this challenge. This study presents two physics-informed neural network (PINN) configurations, PINN-parallel and PINN-series, designed to improve SOH prediction by combining an equivalent circuit model (ECM) with a long short-term memory (LSTM) network. PINN-parallel process inputs data through parallel ECM and LSTM modules and combines their outputs for SOH estimation. On the other hand, the PINN-series uses a sequential approach that feeds ECM-derived parameters into the LSTM network to supplement temporal data analysis with physics information. Both models utilize easily accessible voltage, current, and temperature data that match realistic battery monitoring constraints. Experimental evaluations show that the PINN-series outperforms the PINN-parallel and the baseline LSTM model in accuracy and robustness. It also adapts well to different input conditions. This demonstrates that the simulated battery dynamic states from ECM increase the LSTM's ability to capture degradation patterns and improve the model's ability to explain complex battery behavior. However, a trade-off between the robustness and training efficiency of PINNs is identified. The research outcomes show the potential of PINN models (particularly the PINN-series) in advancing battery management systems, although they require considerable computational resources.

This study aims to introduce an Artificial Intelligence (AI) guided computational framework for the automatic identification, inspection, assessment, and remanufacturing of end-of-use products. The proposed framework consists of three main steps: (1) developing computer vision and image processing algorithms for analyzing product teardown images, (2) quantifying the economic and environmental value of remanufacturing from product images, and (3) developing recommender algorithms to identify the best recovery decision for each device. The study discusses the importance of advancing object detection, image segmentation, and machine learning algorithms to automatically compute the value embedded in discarded items and developing recommendation systems to determine remanufacturing operations from product configurations. The main focus of the study is on the value assessment and remanufacturing of electronic waste (e-waste). The work emphasizes the need for developing object detection for identifying small objects (e.g., screws, bolts, snaps) and overlapped components (e.g., cables, printed circuit boards) standard in the design of consumer electronics by incorporating product shapes and features. The proposed value assessment framework has applications beyond remanufacturing and can be used in take-back programs and other business models that benefit from product serialization and assessment of individual devices.

Robotic technology can benefit disassembly operations by reducing human operators’ workload and assisting them with handling hazardous materials. Safety consideration and predicting human movement is a priority in human-robot close collaboration. The point-by-point forecasting of human hand motion which forecasts one point at each time does not provide enough information on human movement due to errors between the actual movement and predicted value. This study provides a range of possible hand movements to enhance safety. It applies three machine learning techniques including Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Bayesian Neural Network (BNN) combined with Bagging and Monte Carlo Dropout (MCD), namely LSTM-Bagging, GRU-Bagging, and BNN-MCD to predict the possible movement range. The study uses an Inertial Measurement Units (IMU) dataset collected from the disassembly of desktop computers to show the application of the proposed method. The findings reveal that BNN-MCD outperforms other models in forecasting the range of possible hand movement.

Accurate prediction of repair duration is an important challenge in product maintenance due to its implications for resource allocation, customer satisfaction, and operational performance. This study aims to develop a deep learning framework to help fleet repair shops accurately categorize repair time given product historical data. The study uses an automobile repair and maintenance dataset and creates an end-to-end predictive framework by employing a multi-head attention network designed for tabular data. The developed framework combines categorical information, transformed through embeddings and attention mechanisms, with numerical historical data to facilitate integration and learning from diverse data features. A weighted loss function is introduced to overcome class imbalance issues in large datasets. Moreover, an online learning strategy is used for continuous incremental model updates to maintain predictive accuracy in evolving operational environments. Our empirical findings demonstrate that the multi-head attention mechanism extracts meaningful interactions between vehicle identifiers and repair types compared to a feed-forward neural network and a random forest model. Also, combining historical maintenance data with an online learning strategy facilitates real-time adjustments to changing patterns and increases the model's predictive performance on new data. The model is tested on real-world repair data spanning 2013 to 2020 and achieves an accuracy of 78

The increasing volume of electronic waste (e-waste) creates significant environmental and economic challenges which demands practical management strategies. Life Cycle Assessment (LCA) has been known as a principal tool for evaluating the environmental impact of e-waste recycling and disposal methods. However, its application is hampered by inconsistencies in methodology, data limitations, and variations in system boundaries. This study provides a review of current LCA tools used in e-waste analysis and identifies gaps and opportunities for improvement. It categorizes studies into three groups: studies that applied LCA to product and process optimization, impact evaluation, and policy development. Findings reveal that LCA has been helpful in assessing the sustainability of different recycling strategies. However, significant variations exist in methodological approaches and data accuracy. Challenges such as the lack of standardized LCA protocols, the limited availability of region-specific impact data, and inconsistencies in assessment methodologies are still barriers to its widespread adoption. Finally, the study discusses emerging trends in LCA aimed at addressing current gaps, including the incorporation of machine learning and artificial intelligence for predictive modeling, dynamic impact assessment frameworks, and the role of real-time data collection via IoT-based sensors.

Product disassembly is integral to remanufacturing and recovery operations of end-of-use devices. Traditionally, disassembly has been conducted manually with significant safety risks to human workers. In recent years, robotic disassembly has gained popularity to alleviate human workload and safety concerns. Despite these advancements, robots have limited capabilities in handling all disassembly tasks independently. It is essential to assess whether a robot is capable of performing specific disassembly tasks or not. This study proposes a disassembly scoring framework that evaluates robotic feasibility for disassembling components based on five design-related factors: weight, shape, size, accessibility, and positioning. For each factor, a disassembly score is defined to analyze its specific impact on robotic grasping and placement capabilities. Further, the relationship between the five factors and robotic capabilities, such as grasping and placing, is discussed by an example of the UR5e manipulator. To show the potential for automating the generation of disassembly metric, the Multi-Axis Vision Transformer (MaxViT) model is used to determine component sizes through image processing of the XPS 8700 desktop. Moreover, the application of the proposed disassembly scoring framework is discussed in terms of determining the appropriate work setting for disassembly operations under three main categories: human–robot collaboration (HRC), Semi-HRC, and Worker-Only settings. A disassembly time metric for calculating disassembly time for HRC is also proposed. The study outcomes determine the proper work settings based on the robotic capability.

This article investigates the design decision facing original equipment manufacturers (OEMs) in determining the optimal degree of repairability by considering market dynamics. The article develops a game-theoretic model to optimize the degree of product repairability for smartphones in a market in which an OEM and a coalition of independent service providers compete in offering repair services. A survey is conducted to estimate the consumer-related parameters of the game theory model by considering factors such as repair cost, prior repair experience of customers, and the quality of repair services offered by the OEM and independent repair service providers. The findings reveal that regardless of the repairability level, the OEM's repair profits are maximized when a significant disparity in the quality of repair services between the OEM and their competitors exists. On the other hand, independent repair service providers' profits are maximized when there is a low disparity in the quality of repair services. Also, the results show why the adoption of a fully repairable device is not the optimal strategy adopted by OEMs. Instead, a sufficiently large degree of repairability can be the strategic choice, as it maximizes the total OEM's profits derived from both the sale of future products and the provision of repair services for previously sold devices. At the same time, this strategy can encourage repair practices among consumers toward a more sustainable society.

This paper aims to analyze consumer behavior in laptop lithium-ion battery consumption throughout their life cycle. As the demand for battery-powered products continues to grow, helping consumers select batteries that align with their actual usage patterns is important for promoting sustainable consumption. However, there is limited research on whether consumers' choices of battery features, such as capacity, are compatible with their real consumption needs. To investigate this, a multinomial logistic regression model is developed to predict battery health status over time. The study uses a dataset of 719 records collected from student laptop users in Chicago, IL. The dataset includes technical specifications and usage metrics such as charging cycles, full and design capacities, and battery age. The findings show that each additional cycle increases the likelihood of degradation by 0.022. On the other hand, batteries with larger design capacities tend to be more durable, with each additional unit of capacity reducing the likelihood of degradation by 0.0011. Next, an optimized consumption scenario is suggested to demonstrate how aligning battery choice with real usage needs can lead to more sustainable outcomes. The results show a nearly 60% reduction in the likelihood of battery degradation, achieved by better matching battery capacity with consumer' actual needs. Finally, we discuss the sustainability benefits of the proposed scenario.

Human-robot collaboration (HRC) has become an integral element of many manufacturing and service industries. A fundamental requirement for safe HRC is understanding and predicting human trajectories and intentions, especially when humans and robots operate nearby. Although existing research emphasizes predicting human motions or intentions, a key challenge is predicting both human trajectories and intentions simultaneously. This paper addresses this gap by developing a multi-task learning framework consisting of a bi-long short-term memory-based encoder-decoder architecture that obtains the motion data from both human and robot trajectories as inputs and performs two main tasks simultaneously: human trajectory prediction and human intention prediction. The first task predicts human trajectories by reconstructing the motion sequences, while the second task tests two main approaches for intention prediction: supervised learning, specifically a support vector machine, to predict human intention based on the latent representation, and, an unsupervised learning method, the hidden Markov model, that decodes the latent features for human intention prediction. Four encoder designs are evaluated for feature extraction, including interaction-attention, interaction-pooling, interaction-seq2seq, and seq2seq. The framework is validated through a case study of a desktop disassembly task with robots operating at different speeds. The results include evaluating different encoder designs, analyzing the impact of incorporating robot motion into the encoder, and detailed visualizations. The findings show that the proposed framework can accurately predict human trajectories and intentions.

Despite the importance of product repairability, current methods for assessing and grading repairability are limited, which hampers the efforts of designers, remanufacturers, original equipment manufacturers (OEMs), and repair shops. To improve the efficiency of assessing product repairability, this study introduces two artificial intelligence (AI) based approaches. The first approach is a supervised learning framework that utilizes object detection on product teardown images to measure repairability. Transfer learning is employed with machine learning architectures such as ConvNeXt, GoogLeNet, ResNet50, and VGG16 to evaluate repairability scores. The second approach is an unsupervised learning framework that combines feature extraction and cluster learning to identify product design features and group devices with similar designs. It utilizes an oriented FAST and rotated BRIEF feature extractor (ORB) along with k-means clustering to extract features from teardown images and categorize products with similar designs. To demonstrate the application of these assessment approaches, smartphones are used as a case study. The results highlight the potential of artificial intelligence in developing an automated system for assessing and rating product repairability.

Electric vehicles (EVs) are considered an environmentally friendly option to conventional vehicles. As the most critical module in EVs, batteries are complex electrochemical components with nonlinear behavior. On-board battery system performance is also affected by complicated operating environments. Real-time EV battery in-service status prediction is tricky but vital to enable fault diagnosis and aid in the prevention of dangerous occurrences. Data-driven models with advantages in time series analysis can be used to capture the degradation pattern from data about certain performance indicators and predict the battery states. The Transformer model is capable of capturing long-range dependencies efficiently using a multi-head attention block mechanism. This paper presents the implementation of a standard Transformer and an encoder-only Transformer neural network to predict EV battery state of health (SOH). Based on the analysis of the lithium-ion battery from NASA Prognostics Center of Excellence website’s publicly accessible dataset, 28 features related to the charge and discharge measurement data are extracted. The features are screened using Pearson correlation coefficients. The results show that the filtered features can effectively improve the accuracy of the model as well as the computational efficiency. The proposed standard Transformer shows good performance in SOH prediction.