Journal of Intelligent Manufacturing最新文献

Manufacturing systems have recently witnessed a shift from the widely adopted automated systems seen throughout industry. The evolution of Industry 4.0 or Smart Manufacturing has led to the introduction of more autonomous systems focused on fault tolerant and customized production. These systems are required to utilize multimodal data such as machine status, sensory data, and domain knowledge for complex decision making processes. This level of intelligence can allow manufacturing systems to keep up with the ever-changing markets and intricate supply chain. Current manufacturing lines lack these capabilities and fall short of utilizing all generated data. This paper delves into the literature aiming at achieving this level of complexity. Firstly, it introduces cognitive manufacturing as a distinct research domain and proposes a definition by drawing upon various preexisting themes. Secondly, it outlines the capabilities brought forth by cognitive manufacturing, accompanied by an exploration of the associated trends and technologies. This contributes to establishing the foundation for future research in this promising field.

One core concept of Industry 4.0 is establishing highly autonomous manufacturing environments. In the vision of Industry 4.0, the product leads its way autonomously through the shopfloor by communicating with the production assets. Therefore, a common vocabulary and an understanding of the domain’s structure are mandatory, so foundations in the form of knowledge bases that enable autonomous communication have to be present. Here, ontologies are applicable since they define all assets, their properties, and their interconnection of a specific domain in a standardized manner. Reusing and enlarging existing ontologies instead of building new ontologies facilitates cross-domain and cross-company communication. However, the demand for reusing or enlarging existing ontologies of the manufacturing domain is challenging as no comprehensive review of present manufacturing domain ontologies is available. In this contribution, we provide a holistic review of 65 manufacturing ontologies and their classification into different categories. Based on the results, we introduce a priority guideline and a framework to support engineers in finding and reusing existent ontologies of a specific subdomain in manufacturing. Furthermore, we present 16 supporting ontologies to be considered in the ontology development process and eight catalogs that contain ontologies and vocabulary services.

Molds are assembled manually due to a shortage of skilled workers and challenges associated with automating operations, which arise from the low-volume, high-variety characteristics of mold production. This study proposed a human–robot collaborative mold assembly using two heterogeneous collaborative robots to address the ergonomic concerns. The use of two heterogeneous cobots enables the handling of different assembly requirements. The diversity of mold structure and different specifications of resources require comprehensive knowledge management to enable interaction and collaboration among resources. However, knowledge management in the domain of mold assembly is yet to be developed in a format understandable by both human and robots. Therefore, a framework of knowledge management is proposed to manage the knowledge within the human–robot collaboration (HRC) in a mold assembly domain. This framework includes an ontology-based decision making that utilizes outcomes from task assignment to decide the mold parts arrangement within the HRC workspace. A set of rules are modeled in the developed ontology for knowledge reasoning according to the use case of collaborative assembly of two-plate injection mold. In addition to part arrangement, the developed HRC ontology can be used to extract data and information based on user’s request and decisions, such as tool selection for subtask execution. The HRC mold assembly ontology serves as a stepping stone towards developing a context-based decision making for multi-resources HRC in future implementation.

Due to the complexity of electrical components, traditional edge detection methods cannot always accurately extract key edge features of them. Therefore, this study constructs a dataset of complex electrical components and proposes a Step-by-Level Multi-Scale Extraction, Fusion, and Refinement Network (SMFRNet) that is based on the salient object detection algorithm. As detailed features includes a wealth of texture and shape characteristics that are related to edges, so the Hierarchical Deep Aggregation U-block (HDAU) is incorporated in the encoder as a means of capturing more details through hierarchical aggregation. Meanwhile, the proposed Multi-Scale Pyramid Convolutional Fusion (MPCF) and Fusion Attention Structure (FAS) achieve step-by-level feature refinement to obtain finer edges. In order to address the issues of imbalanced pixel categories and the difficulty in separating edge pixels, a hybrid loss function is also constructed. The experimental results indicate that this method outperforms nine state-of-the-art algorithms, enabling the extraction of high-precision key edge features. It provides a reliable method for key edge extraction in complex electrical components and provides important technical support for automated components measurement.

This paper presents a framework for a systematic analysis of the impact of data aggregation on a multi-product multi-period network design problem with batch cost. The optimization objective is to design the vehicle distribution network for an automotive manufacturer. Numerical experiments are conducted with real production data. Given the problem’s scale and complex constraints, data aggregation emerges as a natural strategy to help the convergence of resolution methods towards good solutions. We explore three aggregation dimensions: product type, spatial, and temporal, and for each of them, different levels. Addressing multiple aggregation dimensions is a novel approach that has not been extensively explored in current literature, especially within industrial settings. Our aggregation-disaggregation method reveals that data aggregation consistently leads to improved solutions within a constrained computation time, with temporal aggregation demonstrating the most significant reduction in problem size and solution improvement. Lastly, we give some managerial insights considering the industrial context.

Assembly is an essential stage in industrial electronic equipment manufacturing and needs to meet the complexity of manufacturing. Therefore, the assembly process planning for industrial electronic equipment still relies on the experiences of planners. The advent of knowledge graphs brings an opportunity to achieve automated assembly process planning. Thus, extracting process knowledge from historical assembly process documents and constructing assembly process knowledge graphs are indispensable. However, the complexity of industrial electronic equipment manufacturing leads to assembly process documents containing more complex assembly relations, longer texts, and high-density assembly entities. These characteristics pose challenges to assembly process knowledge extraction and knowledge graph modeling. The confidentiality of assembly process documents further hinders the development of this field. To address these challenges, we propose a pipeline for achieving assembly process planning from historical assembly process documents. First, we construct an assembly process dataset using historical assembly process documents from an industrial electronic equipment enterprise. Then, we propose a global relation-driven bidirectional extraction model, which automatically constructs the assembly process knowledge graph. In addition, we also propose a knowledge graph-based matching and searching method to support process planning. The proposed model is evaluated on the constructed dataset and a publicly accessible equipment fault diagnostic dataset, achieving F1-scores of 92.9% and 87.9%, respectively. Experimental results demonstrate that the proposed model achieves state-of-the-art performance on both datasets. Furthermore, we construct an assembly process knowledge graph for industrial electronic equipment and perform assembly process planning, which validates the feasibility of our pipeline.

The real-time and accurate prediction of the stress–strain and deformations field of material is a vital function for the intelligent press fitting system of the rubber bushing. The physics-informed neural network (PINN) provide an efficient approach to constructing physical fields with high robustness and interpretability in real time. However, currently, PINN usually solves problems under known boundary conditions, which are not given explicitly in most realistic engineering problems. This study proposes a cascaded multitask PINN (CM-PINN) that divides the problem solving of rubber bushing interference fit into two phases: boundary computation and forward solving of the physical field. In CM-PINN, one sub-network is used for boundary computation, while two other sub-networks are used for computing the physical fields of hyperelastic material, rubber. In both stages, physical constraints are incorporated into the sub-networks. These subnetworks are trained hybridly through the cascaded framework using data obtained from the finite element model (FEM), which was verified by experimental results. In order to validate the CM-PINN model, FEM data are used as a reference solution for comparison with conventional PINN. To evaluate the advantages of CM-PINN, ablation tests are conducted by randomly selecting training samples with different sizes. It is found that CM-PINN has higher accuracy and convergence compared to hybrid output PINNs. CM-PINN shows remarkable improvement in its generalization ability in the case of small sample size, underscoring its robust applicability across different data scenarios.

Digital twin (DT) is an important approach for the factory to achieve intelligence. Due to the different scenarios and definitions, the generalization of frameworks for DT-based smart factories is weak, slowing down the overall process of industrial intelligence. Meanwhile, the pressure of data transmission and processing increases dramatically because of data explosion, which poses a challenge to the rational allocation of computing resources. In addition, more advanced strategies for training and running models are needed to support more sophisticated services. This paper proposes a reference framework that combines DT and cloud-fog-edge computing collaboration (CFE). First, the DT fuses physical and virtual spaces. The virtual-real fusion provides more information for operations, and the virtual space gives more accurate and timely decisions based on the constantly refreshed state. Secondly, by introducing CFE, suitable operating platforms for each layer of the DT-based smart factory are set, which enhances data interaction and reduces the dependence on cloud computing. The DT-CFE framework is well generalized. This paper first introduces the definition of the DT-based smart factory and its components. Then the methodology of the DT-CFE-based smart factory is proposed, and the network topology and operation mechanism are introduced. In this framework, the transmission and response performance of its data interaction is tested, and the interference of dynamic events occurring through scheduling is studied to illustrate the effectiveness and superiority of the framework.

High-precision optical products made of polymeric materials have been surging in recent years due to the prevalence of smartphones and their camera modules. Manufacturing fast-changing generations of high-precision optical lenses with accurately predicted qualities is a challenging task. Simulations and modern artificial intelligence (AI) techniques play crucial roles in accelerating precise process development. Coupled with computer simulation, this research employs a fusion of explainable AI (XAI) and multi-stage transfer learning (TL) approaches with artificial neural network (ANN) models to predict the surface profile variation of injection-molded polycarbonate (PC) lenses. The proposed method efficiently bridges preliminary simulations to injection molding experiments, covering a complete process development workflow from feature selection, process modeling, to experimental investigation in the same modeling domain. Only one model from scratch is required, which carries knowledge to the final quality prediction model. When compared with the conventional TL and the naïve model, the multi-stage TL approach provides better predictions with a maximum reduction of 64% and 43% in simulation and actual manufacturing data requirement, respectively. This research demonstrates a viable connection between each stage in the injection molding (IM) process development in predicting the qualities of high-precision optical lenses. Meanwhile, the combined usage of XAI and (multi-stage) TL confirms model explanations and pinpoints a potential pathway to assess future TL capabilities from the modeling perspectives.

Process improvement initiatives require access to frequently updated and good quality data. This is an extremely difficult task in the area of production processes, where the lack of a process digital footprint is a very big challenge. To solve this problem, the authors of this article designed, implemented, and verified the results of a new work measurement method. The Workplace Performance Measurement (WPM) method is focused not only on the measurement of task duration and frequency, but also on searching for potential anomalies and their reasons. The WPM method collects a wide range of workspace parameters, including workers' activities, workers' physiological parameters, and tool usage. An application of Process Mining and Machine Learning solutions has allowed us to not only significantly increase the quality of analysis (compared to analog work sampling methods), but also to implement an automated controlling solution. The genuine value of the WPM is attested to by the achieved results, like increased efficiency of production processes, better visibility of process flow, or delivery of input data to MES solutions. MES systems require good quality, frequently updated information, and this is the role played by the WPM, which can provide this type of data for Master Data as well as for Production Orders. The presented authorial WPM method reduces the gap in available scholarship and practical solutions, enabling the collection of reliable data on the actual flow of business processes without their disruption, relevant for i.a. advanced systems using AI.