https://www.mdu.se/

There is an increasing interest from manufacturing industry and academia to improve operational efficiency with advanced AI systems. One such innovation is the Retrieval-Augmented Generation (RAG) system, which integrates a Large Language Model (LLM) to deliver customized recommendations and question-answering based on historical maintenance data and machine operational manuals. Previous studies highlight the effectiveness of LLMs in processing unstructured text, but limited research exists on their application within RAG frameworks for industrial maintenance use cases. The key challenge is efficiently retrieving relevant quality data and providing recommendations that are both contextually appropriate and implementable by maintenance repairmen and technicians. This paper presents a case study where a RAG system, trained on data from a Computerized Maintenance Management System and one operational manual provide recommendations/question-answers to support decision-making on the factory floor. The methodology involves embedding all the data using LLM, followed by conducting a similarity search to identify relevant information. Recommendations are then generated using the LLM and subsequently validated through expert review by subject matter experts. The results indicate that the RAG system facilitates faster and more streamlined decision-making in maintenance tasks by efficiently retrieving and contextualizing contextual maintenance records. This approach not only makes it easier for operators, repairmen and technicians to access expert knowledge, but it also shows how RAG systems can help simplify operations in industries that depend on complicated maintenance processes. The study highlights how RAG-based LLM systems can improve maintenance work by using AI to provide helpful insights.

Navigating spaces without GPS signal access requires drones to rely on an internal 3D representation of the space to navigate in including object of interest and relevance. Many environments contain dynamic objects: doors may close, objects may topple, and additional dimensions, such as heat or limited visibility that become essential to represent. Such comprehensive modeling allows drones to avoiding hazards when executing a traversal from point A to point B within the space or searching for e.g. injured humans needing help or rescue. Sharing spatial information between collaborating drones and human operators, as well as dynamically and jointly updating the 3D model in real-time, introduces specific requirements for these representations and the model must support updates from both drones and human operators. In this paper, we systematically review a range of 3D model representations, classifying them based on applicability to Search and Rescue (SAR) scenarios. The purpose of this literature review is to aid in selecting a suitable 3D representation for SAR operations.

Unmanned Aerial Vehicles (UAVs) are increasingly crucial in Search and Rescue (SAR) operations due to their ability to enhance efficiency and reduce costs. Search and Rescue is a vital activity as it directly impacts the preservation of life and safety in critical situations, such as locating and rescuing individuals in perilous or remote environments. However, the effectiveness of these operations heavily depends on the quality of sensor data for accurate target detection. This study investigates the application of the Real Enhanced Super-Resolution Generative Adversarial Networks (Real-ESRGAN) algorithm to enhance the resolution and detail of infrared images captured by UAV sensors. By improving image quality through super-resolution, we then assess the performance of the YOLOv8 target detection algorithm on these enhanced images. Preliminary results indicate that Real-ESRGAN significantly improves the quality of low-resolution infrared data, even when using pre-trained models not specifically tailored to our dataset, this highlights a considerable potential of applying the algorithm in the preprocessing stages of images generated by UAVs for search and rescue operations.

Ontology development plays a vital role as it provides a structured way to represent and organize knowledge. It has the potential to connect and integrate data from different sources, enabling a new class of AI-based services and systems such as decision support systems and recommender systems. However, in large manufacturing industries, the development of such ontology can be challenging. This paper presents a use case of an application ontology development based on machine breakdown work orders coming from a Computerized Maintenance Management System (CMMS). Here, the ontology is developed using a Knowledge Meta Process: Methodology for Ontology-based Knowledge Management. This ontology development methodology involves steps such as feasibility study, requirement specification, identifying relevant concepts and relationships, selecting appropriate ontology languages and tools, and evaluating the resulting ontology. Additionally, this ontology is developed using an iterative process and in close collaboration with domain experts, which can help to ensure that the resulting ontology is accurate, complete, and useful for the intended application. The developed ontology can be shared and reused across different AI systems within the organization, facilitating interoperability and collaboration between them. Overall, having a well-defined ontology is critical for enabling AI systems to effectively process and understand information.

Research on unmanned autonomous vehicles (UAVs) for search and rescue (SAR) missions is widespread due to its cost-effectiveness and enhancement of security and flexibility in operations. However, a significant challenge arises from the quality of sensors, terrain variability, noise, and the sizes of targets in the images and videos taken by them. Generative adversarial networks (GANs), introduced by Ian Goodfellow, among their variations, can offer excellent solutions for improving the quality of sensors, regarding super-resolution, noise removal, and other image processing issues. To identify new insights and guidance on how to apply GANs to detect living beings in SAR operations, a PRISMA-oriented systematic literature review was conducted to analyze primary studies that explore the usage of GANs for edge or object detection in images captured by drones. The results demonstrate the utilization of GAN algorithms in the realm of image enhancement for object detection, along with the metrics employed for tool validation. These findings provide insights on how to apply or modify them to aid in target identification during search stages.

Companies must enhance total maintenance effectiveness to stay competitive, focusing on both digitalization and basic maintenance procedures. Digitalization offers technologies for data-driven decision-making, but many maintenance decisions still lack a factual basis. Prioritizing efficiency and effectiveness require analyzing equipment history, facilitated by using Computerized Maintenance Management Systems (CMMS). However, CMMS data often contains unstructured free-text, leading to manual analysis, which is resource-intensive and reactive, focusing on short time periods and specific equipment. Two approaches are available to solve the issue: minimizing free-text entries or using advanced methods for processing them. Free-text allows detailed descriptions but may lack completeness, while structured reporting aids automated analysis but may limit fault description richness. As knowledge and experience are vital assets for companies this research uses a hybrid approach by combining Natural Language Processing with domain specific ontology and Large Language Models to extract information from free-text entries, enabling the possibility of real-time analysis e.g., identifying recurring failure and knowledge sharing across global sites.

Unmanned Aerial Vehicle (UAV) swarms, also known as drone swarms, have been a subject of extensive research due to their potential to enhance monitoring, surveillance, and search missions. Coordinating several drones flying simultaneously presents a challenge in increasing their level of automation and intelligence to improve strategic organization. To address this challenge, we propose a solution that uses hill climbing, potential fields, and search strategies in conjunction with a probability map to coordinate a UAV swarm. The UAVs are autonomous and equipped with distributed intelligence to facilitate a cooperative search application. Our results show the effectiveness of the swarm, indicating that this approach is a promising approach to addressing this problem.

Search missions for objects are relevant in both industrial and civilian context, such as searching for a missing child in a forest or to locating equipment in a building or large factory. To send out a drone swarm to quickly locate a misplaced item in a factory, a missing machine on a building site or a missing child in a forest is very similar. Image-based Machine Learning algorithms are now so powerful that they can be trained to identify objects with high accuracy in real time. The next challenge is to perform the search as efficiently as possible, using as little time and energy as possible. If we have information about the area to search, we can use heuristic and probabilistic methods to perform an efficient search. In this paper, we present a case study where we developed a method and approach to evaluate different search algorithms enabling the selection of the most suitable, i.e., most efficient search algorithm for the task at hand. A couple of probabilistic and heuristic search methods were implemented for testing purposes, and they are the following: Bayesian Search together with a Hill Climbing search algorithm and Bayesian Search together with an A-star search algorithm. A swarm adapted lawn mower search strategy is also implemented. In our case study, we see that the performance of the search heavily depends on the area to search in and domain knowledge, e.g., knowledge about how a child is expected to move through a forest area when lost. In our tests, we see that there are significant gains to be made by selecting a search algorithm suitable for the search context at hand.

There is an increased need for protection against unauthorized entry of drones as there has been an increased number of reports of UAV’s entering restricted areas. In this paper we explore an approach of using a swarm/team of drones that are able to cooperate, to autonomously engage and disable one or more unauthorized drones entering a restricted area. In our approach, we have investigated technologies for distributed decision-making and task allocation in real-time, in a dynamic simulated environment and developed descriptive models for how such technologies may be exploited in a mission designed for a drone swarm. This includes the definition of discrete tasks, how they interact and how they are composed to form such a mission, as well as the realization and execution of these tasks using machine learning models combined with behaviour trees. To evaluate our approach, we use a simulated environment for mission execution where relevant KPI’s related to the design of the mission have been used to measure how efficient our approach is in deterring or incapacitating unauthorized drones. The evaluation has been performed using Monte-Carlo simulations on a batch of randomized scenarios and measures of effectiveness has been used to measure each scenario instance and later compiled into a final assessment for the main scenario as well as each ingoing task. The results show a mission success in 93% of the simulated scenarios. Of these 93%, 58% of the scenarios resulted in the threat being neutralized and in 35% of the scenarios the threat was driven away from the critical area. We believe that the application of such measurements aids to validate the applicability of this capability in a real-world scenario and in order to assert the relevance of these parameters, future validations in real-world operational scenarios are warranted.

For the last few years, cases of applying artificial intelligence (AI) to engineering activities towards sustainability have been reported. Life Cycle Engineering (LCE) provides a potential to systematically reach higher and productivity levels, owing to its holistic perspective and consideration of economic and environmental targets. To address the current gap to more systematic deployment of AI with LCE (AI-LCE) we have performed a systematic literature review emphasizing the three aspects:(1) the most prevalent AI techniques, (2) the current AI-improved LCE subfields and (3) the subfields with highly enhanced by AI. A specific set of inclusion and exclusion criteria were used to identify and select academic papers from several fields, i.e. production, logistics, marketing and supply chain and after the selection process described in the paper we ended up with 42 scientific papers. The study and analysis show that there are many AI-LCE papers addressing Sustainable Development Goals mainly addressing: Industry, Innovation, and Infrastructure; Sustainable Cities and Communities; and Responsible Consumption and Production. Overall, the papers give a picture of diverse AI techniques used in LCE. Production design and Maintenance and Repair are the top explored LCE subfields whereas logistics and Procurement are the least explored subareas. Research in AI-LCE is concentrated in a few dominating countries and especially countries with a strong research funding and focus on Industry 4.0; Germany is standing out with numbers of publications. The in-depth analysis of selected and relevant scientific papers are helpful in getting a more correct picture of the area which enables a more systematic approach to AI-LCE in the future.