Computers in Industry最新文献

Efficient performance monitoring in production systems holds paramount importance as it enables organizations to optimize their manufacturing processes, enhance productivity, and maintain a competitive edge in the market. Typically, machine and system level performance monitoring systems are investigated independently, whereas an integrated approach that considers both levels can offer valuable insights and benefits. This paper introduces a data-driven approach for evaluating and improving the performance of production lines by monitoring the performance of both individual machines and their interactions as a system. The approach begins with a rigorous methodology for classifying machine states recorded by the Manufacturing Execution System (MES) into finer-grained substates, enabling a comprehensive analysis of machine cycle time variability. Subsequently, these substates are leveraged as a foundation for constructing performance monitoring models at both the machine and system levels, employing probabilistic automata for the machine level and logistic regression for the system level. The system-level performance monitoring model is constructed to predict a Flow metric that enables the prediction of abnormal behaviors and deviations from production targets. This data-driven approach serves as a foundational ingredient of a system-level digital twin, designed to provide production lines with insights that enable proactive implementation of measures aimed at optimizing overall manufacturing efficiency. Through an industrial test case from the automotive industry, the results demonstrate the capability of performance monitoring, capturing errors within confidence intervals, and establishing predictive cause-and-effect relationships between machines within the production system.

Contract review is an essential step in construction projects to prevent potential losses. However, the current methods for reviewing construction contracts lack effectiveness and reliability, leading to time-consuming and error-prone processes. Although large language models (LLMs) have shown promise in revolutionizing natural language processing (NLP) tasks, they struggle with domain-specific knowledge and addressing specialized issues. This paper presents a novel approach that leverages LLMs with construction contract knowledge to emulate the process of contract review by human experts. Our tuning-free approach incorporates construction contract domain knowledge to enhance language models for identifying construction contract risks. The use of natural language when building the domain knowledge base facilitates practical implementation. We evaluated our method on real construction contracts and achieved solid performance. Additionally, we investigated how LLMs employ logical thinking during the task and provided insights and recommendations for future research.

Maintenance records in Computerized Maintenance Management Systems (CMMS) contain valuable human knowledge on maintenance activities. These records primarily consist of noisy and unstructured texts written by maintenance experts. The technical nature of the text, combined with a concise writing style and frequent use of abbreviations, makes it difficult to be processed through classical Natural Language Processing (NLP) pipelines. Due to these complexities, this text must be normalized before feeding to classical machine learning models. Developing these custom normalization pipelines requires manual labor and domain expertise and is a time-consuming process that demands constant updates. This leads to the under-utilization of this valuable source of information to generate insights to help with maintenance decision support. This study proposes a Technical Language Processing (TLP) pipeline for semantic search in industrial text using BERT (Bidirectional Encoder Representations), a transformer-based Large Language Model (LLM). The proposed pipeline can automatically process complex unstructured industrial text and does not require custom preprocessing. To adapt the BERT model for the target domain, three unsupervised domain fine-tuning techniques are compared to identify the best strategy for leveraging available tacit knowledge in industrial text. The proposed approach is validated on two industrial maintenance records from the mining and aviation domains. Semantic search results are analyzed from a quantitative and qualitative perspective. Analysis shows that TSDAE, a state-of-the-art unsupervised domain fine-tuning technique, can efficiently identify intricate patterns in the industrial text regardless of associated complexities. BERT model fine-tuned with TSDAE on industrial text achieved a precision of 0.94 and 0.97 for mining excavators and aviation maintenance records, respectively.

In recent years, deep learning-based approaches for industrial surface defect detection have shown great promise. To address the domain shift issue among data from different sources in the industrial domain, we present a novel plug-and-play Style Adaptation (SA) module, which endows the equipped defect detector with the capability to exhibit robustness to diverse styles present within the samples. This module effectively leverages datasets sourced from diverse origins while possessing congruent data types. In contrast to other domain adaptation approaches lacking well-defined domain delineations, the SA module generates representations characterized by distinct practical implications and precise mathematical formulations. Moreover, incorporating attention mechanisms reduces the need for manual intervention, allowing the module to focus autonomously on crucial branches in it. Experimental results demonstrate the superior efficacy of our approach compared to state-of-the-art techniques. Furthermore, an authentic dataset from various manufacturers is publicly available for deep learning research and industrial applications. Access the dataset at: https://github.com/THU-PMVAI/MTS3D

BizDevOps is an emerging trend that seeks to cut back the lag between product/service vision and implementation. However, so far this trend has been mainly unnoticed by research. This paper carries out a “grey literature” (non-academic) review on BizDevOps. Data is collected from reports, articles, webpages, and blog posts to capture the professionals’ insights on BizDevOps. We develop a conceptual framework for BizDevOps that organizes and integrates concepts and constructs embedded in the grey literature. Based on this, the paper offers insights for organizations aiming to move towards the BizDevOps approach and identifies research opportunities in the BizDevOps area.

Additive Manufacturing (AM), particularly Selective Laser Melting (SLM), has revolutionized the industrial manufacturing sector owing to its remarkable design flexibility and precision. However, it is well known that slight changes in SLM process parameters may highly affect the surface quality of the as-built product. In this paper, we investigate the influence of SLM printing parameters (laser power, laser scanning speed, layer thickness, and hatch distance) on surface quality and develop a predictive model for surface quality based on the given printing parameters. The developed model is constructed by a Bayesian Optimization and soft Attention mechanism-enhanced Transfer learning (BOAT) framework with superior domain adaptability and generalization capability. Through experimental validation, the effectiveness of the BOAT approach in estimating printing parameters and correlating them with surface quality has been verified. The comprehensive methodology, experimental configurations, prediction results, and ensuing discussions are all presented. This study contributes to providing valuable insights and practical implications for improving the competitiveness and impact of SLM in advanced manufacturing by accurately predicting surface quality with specified printing parameters.

While the adoption of open Building Information Modeling (open BIM) standards continues to grow, the inherent complexity and multifaceted nature of the built asset lifecycle data present a critical bottleneck for effective information retrieval. To address this challenge, the research community has started to investigate advanced natural language-based search for building information models. However, the accelerated pace of advancements in deep learning-based natural language processing research has introduced a complex landscape for domain-specific applications, making it challenging to navigate through various design choices that accommodate an effective balance between prediction accuracy and the accompanying computational costs. This study focuses on the semantic tagging of user queries, which is a cardinal task for the identification and classification of references related to building entities and their specific descriptors. To foster adaptability across various applications and disciplines, a semantic annotation scheme is introduced that is firmly rooted in the Industry Foundation Classes (IFC) schema. By taking a comparative approach, we conducted a series of experiments to identify the strengths and weaknesses of traditional and emergent deep learning architectures for the task at hand. Our findings underscore the critical importance of domain-specific and context-dependent embedding learning for the effective extraction of building entities and their respective descriptions.

The current practice for creating as-built geometric Digital Twins (gDTs) of industrial facilities is both labour-intensive and error-prone. In aged industries it typically involves manually crafting a CAD or BIM model from a point cloud collected using terrestrial laser scanners. Recent advances within deep learning (DL) offer the possibility to automate semantic and instance segmentation of point clouds, contributing to a more efficient modelling process. DL networks, however, are data-intensive, requiring large domain-specific datasets. Producing labelled point cloud datasets involves considerable manual labour, and in the industrial domain no open-source instance segmentation dataset exists. We propose a semi-automatic workflow leveraging object descriptions contained in existing gDTs to efficiently create semantic- and instance-labelled point cloud datasets. To prove the efficiency of our workflow, we apply it to two separate areas of a gas processing plant covering a total of $40000{\text{m}}^2$. We record the effort needed to process one of the areas, labelling a total of 260 million points in 70 h. When benchmarking on a state-of-the-art 3D instance segmentation network, the additional data from the 70-hour effort raises mIoU from 24.4% to 44.4%, AP from 19.7% to 52.5% and RC from 45.9% to 76.7% respectively.

In the new hyper connected factories, data gathering, and prediction models are key to keeping both productivity and piece quality. This paper presents a software platform that monitors and detects outliers in an industrial manufacturing process using scalable software tools. The platform collects data from a machine, processes it, and displays visualizations in a dashboard along with the results. A statistical method is used to detect outliers in the manufacturing process. The performance of the platform is assessed in two ways: firstly by monitoring a five-axis milling machine and secondly, using simulated tests. Former tests prove the suitability of the platform and reveal the issues that arise in a real environment, and latter tests prove the scalability of the platform with higher data processing needs than the previous ones.

The data-driven method based on acoustic emission signals is gradually becoming a hot topic in the field of laser shock peening quality monitoring. Although some existing deep learning methods do provide excellent monitoring accuracy and speed, they lack physical interpretability in nature, and the opacity of these decisions poses a great challenge to their credibility. The weak interpretability of deep learning models has become the biggest obstacle to the landing of artificial intelligence projects. To overcome this drawback, this paper proposes a monitoring strategy that can achieve physical interpretability in feature extraction, selection and classification, namely, jointly generating monitoring results and explanations. Specifically, it is an end-to-end model that combines convolutional neural units, gated recurrent units, and attention mechanisms. Firstly, a wavelet analysis with physical meaning that can be autonomously learned is performed on the acoustic emission. Then, the contribution of features is distinguished based on the correlation of information in different frequency bands, and redundant and noisy features are removed. Finally, the interpretability evaluation of processing quality is realized by using gated recurrent units with attention mechanisms. The effectiveness and reliability of the proposed method are confirmed by the experimental data of both laser shock peening at small and large gradient energies compared to state-of-the-art feature methods, CNN- and LSTM-based models. Most importantly, the physical interpretation of acoustic emission signals during the processing can increase the credibility of decisions and provide a basic logic for on-site judgments by professionals.