1,720,963 research outputs found
Advancing Spoken Dialog Systems for Manufacturing: From Conceptual Architecture and Taxonomy to Real Case Applications and Future Directions
No full textThis research encompasses a comprehensive exploration of Spoken Dialogue Systems (SDSs) in the manufacturing sector. It begins by establishing a conceptual architecture and taxonomy to guide the design and selection of SDS elements. Real case applications, including worker safety and cybersecurity support, validate the research findings and highlight areas for improvement. Looking ahead, the study delves into the potential of Large Language Models (LLMs) and multi-modal applications. Emphasizing the importance of extreme personalization, the study highlights the need to cater to the diverse qualifications and preferences of workers. Additionally, it investigates the integration of SDSs with other sensory modalities, such as images, videos, and augmented or virtual reality scenarios, to enhance the user experience and productivity. The research also addresses crucial considerations related to knowledge base optimization. It examines semantic variations of words across different application contexts, the continuous updating of procedures and data, and the adaptability of SDSs to diverse dialects and linguistic abilities, particularly in low-schooling personnel scenarios. Privacy, industrial protection, and ethical concerns in the era of LLMs and external players like OpenAI are given due attention. The study explores the boundaries of knowledge that conversational systems should possess, advocating for transparency, explainability, and responsible data handling practices
Natural Language Processing applications in manufacturing: a systematic literature review
No full textAmong the manufacturing sector several applications of Natural Language Processing (NLP) are emerging. NLP is a branch of Artificial Intelligence (AI) aimed at understanding, interpreting, and manipulating human language through computer-based data processing. This application is quite powerful and prospective in manufacturing context, considering the ever-increasing amount of data available within the organizations, often unstructured, non-standardized, and free text. Therefore, human analysis to extract information and useful knowledge results in a long and tedious task with limited added value. The automation of these activities moves workers to more meaningful and value-added activities; it improves efficiency in searching for and extracting information, with benefits for decision-making processes. The paper presents a systematic literature review concerning NLP applications in manufacturing, conducted using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement methodology. Basing on the documents retrieved, a comparative analysis of the literature is presented. The analysis is carried out following two different rationales: an objective analysis, which highlights and compares the different purposes with which NLP is applied in the manufacturing field, such as knowledge base, ontology, predictive maintenance, human machine interaction and decision support system. The second analysis investigates NLP applications by exploring different production process phases involved in manufacturing activities. The research identified mature NLP applications, transversally implemented in several production process phases, with specific objectives. The paper provides a comprehensive and in-depth overview on the topic. Finally, possible future directions of development of NLP in manufacturing were defined. © 2022, AIDI - Italian Association of Industrial Operations Professors. All rights reserved
Detecting dangerous behaviours and promoting safety in manufacturing using computer vision
Full text linkThe safe operation of forklifts in manufacturing environments is critical for the efficient transportation of goods. However, accidents can occur due to distraction, failure to use Personal Protective Equipment (PPE), and improper handling of the forklift. To improve safety, the authors propose a computer vision solution to monitor forklift operators and their compliance with safety regulations. The model is trained to detect behaviours that could lead to accidents and alert the operator in real time. The proposed solution can be integrated with the forklift's control system, providing immediate feedback to the operator to reduce the risk of accidents. The model uses transfer learning, a technique that leverages pre-trained models to improve the accuracy of the model with limited data. The PoseNet pre-trained model was fine-tuned on a dataset of annotated videos of forklift operators to improve its accuracy in classifying different behaviours. Future work can investigate the integration of the solution with other safety systems to provide a comprehensive safety solution in manufacturing environments
Using Natural Language Processing to uncover main topics in defect recognition literature
No full textThe issue of defect detection is particularly important namely in plant engineering, where it is crucial to ensure high-quality production by minimizing the number of defective parts. In the last years, the interest in the subject has grown a lot and the methods and approaches proposed for defect recognition are multiple. Therefore, when dealing with defect recognition researchers are faced with an increasing number of articles that slows them down in identifying the set of articles of their interest. This work aims to provide a baseline classification of articles based on emerging issues such as the investigated material, the production typology in which the material is included, and the type of analysis to be effected. For these reasons, the paper proposes an automatic solution based on text mining techniques. Specifically, the study applies Natural Language Processing (NLP) to articles’ titles, abstracts, and keywords using two different approaches: K-Means clustering algorithm and Latent Dirichlet Allocation (LDA). K-Means is used to cluster the collection of documents into related groups based on the contents of the particular documents. LDA instead is used to classify the papers using the concept of topic modeling. Articles have been collected from Scopus database. The scope of the research is limited to journal and conference articles, published in English excluding articles classified as reviews, as well as book chapters, books, notes, erratum
Clustering application for condition-based maintenance in time-varying processes: a review using latent dirichlet allocation
Full text linkIn the field of industrial process monitoring, scholars and practitioners are increasing interest in time-varying processes, where different phases are implemented within an unknown time frame. The measurement of process parameters could inform about the health state of the production assets, or products, but only if the measured parameters are coupled with the specific phase identification. A combination of values could be common for one phase and uncommon for another phase; thus, the same combination of values shows a high or low probability depending on the specific phase. The automatic identification of the production phase usually relies on clustering techniques. This is largely due to the difficulty of finding training fault data for supervised models. With these two considerations in mind, this contribution proposes the Latent Dirichlet Allocation as a natural language-processing technique for reviewing the topic of clustering applied in time-varying contexts, in the maintenance field. Thus, the paper presents this innovative methodology to analyze this specific research fields, presenting the step-by-step application and its results, with an overview of the theme
MARLIN Method: Enhancing Warehouse Resilience in Response to Disruptions
Full text linkBackground: Endogenous and exogenous factors impact the operational characteristics of supply chains, affecting wholesale warehouses. The survival of a warehouse is often threatened by disruptive events that alter infrastructure and performance. The emergence of COVID-19 exemplified the need for adaptability in retail goods supply chains, emphasizing the necessity for responding to external shocks. Methods: The MARLIN (Method wArehouse ResiLience dIstruptioN) method, founded on theories and models of resilience engineering is introduced. MARLIN is a practical tool designed to identify key areas requiring intervention in response to disruptive events. An empirical test was conducted in an Italian warehouse. Results: The conducted test yielded tangible results, demonstrating the efficacy of the method. It successfully pinpointed areas necessitating intervention and identified Key Performance Indicators (KPIs) associated with disruptions. The study not only underscores the importance of data collection but also highlights the often-overlooked significance of warehouse management. Conclusions: The study establishes MARLIN as a valuable asset for stakeholders involved in disruption management. Its application has proven instrumental in recognizing areas of intervention and identifying KPIs related to disruptions. Ongoing research endeavors to broaden its applicability across diverse supply chain scenarios, aiming to enhance situational awareness and enable proactive risk assessment through what-if analysis
The Effectiveness of Outsourcing Cybersecurity Practices: A Study of the Italian Context
No full textThe increasing number of cyber-attacks requires an organizational awareness about the disruptive effects of fraud attempts and acts of vandalism on business continuity and, sometimes, on company survival. The context influences the way companies use and adapt these theories in practice, so we consider in this study differences in the effectiveness of cybersecurity best practices between organizations that manage internally or outsource the cybersecurity processes. We conducted a study involving 153 managers’ experts in cybersecurity who responded to a survey on the effectiveness of NIST procedures. Results revealed significant differences in the effectiveness of managing cybersecurity in-house or outsource it. Specifically, major differences can be observed in the variables related to the use of disciplinary processes, the protection of log information, and the use of lessons learned to improve recovery plans. These differences provide further insights for cybersecurity management literature and a practical instrument for organizations willing to adapt their cyber processes to their organizational context
Organizational learning for cybersecurity
No full textThe increasing connectivity and digitization of organizations have made cybersecurity a top priority. Organizations have become highly dependent on integrated systems and data, exposing them to cyber threats that can lead to economic and reputational losses. The COVID-19 pandemic has further highlighted vulnerabilities in cybersecurity systems across sectors. In particular, there has been a recent focus on the role humans play in this scenario by turning out to be both a possible source of vulnerability and a mitigating factor. To manage this, organizations must focus on establishing a culture of cyber security awareness by promoting policies, standards, and users’ behaviors from an organizational learning perspective.
Employees’ education is one of the components needed to create such a culture, hence training programs on cybersecurity have a crucial role.
However, the effectiveness and sustainability of training programs depend on including a variety of stakeholder groups to identify and mitigate cost and efficacy concerns, adopt accessible training techniques, employ trainers with relevant expertise, and address psychological obstacles like trainee guilt and shame. A balanced human-machine approach is needed to maximize the benefits of connectivity while minimizing cyber risks. Overall, cybersecurity training requires an ongoing, collaborative, and flexible process tailored to each organization's context. Today, there is no consensus on the most effective and appropriate cybersecurity training methods. This research investigates available cybersecurity awareness types of training and provides guidelines for developing good organizational training programs in increasingly digital environments
Yet Another Warehouse KPI’s Collection
Full text linkWarehouses are strategic systems for all supply chains since their performances impact operations and efficiency of all direct and indirect stakeholders. Therefore, monitoring warehouses' performances constantly and real-time is getting so important, both to guarantee an effective warehouse management and to detect in advance anomalous and potentially destructive trends. The current literature about warehousing Key Performance Indicators (KPI) appears to lack an extensive collection. Classification logics are often partial or based on specific contexts. At the same time, the amount and typology of data collected on the warehouse often hinder a consistent performance monitoring. This paper aims to fill such gap and guide organizations in identifying the relevant information to gather for warehouse performance monitoring. Firstly, a scoping literature review was conducted to provide an extensive list of warehouse KPIs. Then, the collected results set the groundwork for a dynamic and interactive database called YAWKC. This tool is designed as a knowledge graph allowing for non-linear exploration of data and for continuous enrichment by experts’ contribution, representing the starting point for further knowledge generation in an explorable, dynamic and potentially ever-growing way
Machine learning models to predict components decay in a naval propulsion system
No full textThe decay of a single component in a naval propulsion system can affect the performance of the entire system, involving expensive maintenance costs for restoring efficient conditions. Therefore, a regular control of the decay of key components of these systems is essential for properly handle maintenance actions. Moreover, in naval propulsion systems it is necessary to consider the difficulty in implementing an onboard maintenance action or returning a vessel. Two relevant components in naval propulsion systems are the turbine and the compressor. This study develops two machine learning models to predict turbine and compressor decay, i.e. based on classification and regression approaches. The former classifies whether the components are decayed or not, thus highlighting a state of criticality, the latter predicts a specific value of each decay coefficient. For each approach, different algorithms are compared, e.g. boosted trees, linear regression or support vector machines. A case study considering sixteen inputs has been used to test the effectiveness of the proposed solution, starting from a dataset of about twelve thousand instances referred to a naval vessel. A sensitivity analysis of relevant parameters has been developed to verify the robustness of the approach
- …
