The concept of digital factories is becoming more popular among the manufactures due to its contributions in simplifying the work of plant engineers and reducing time to market of the products [1]. The main aim of this thesis was to investigate the possible applications of a digital copy of production processes by creating a virtual model of a seam sealing line in the paint department at Scania in Oskarshamn. The main usage of the virtual model was to perform offline programming of the robots. The benefits of offline programming were investigated and compared to online programming. Some comparisons between offline and online programming were made when it comes to the time needed for programming, the accuracy of generated robot paths, the workflow of both approaches and the required knowledge to turn from online programming to offline programming.

The result of the conducted experiment showed that the time of robot programming can be reduced, and the accuracy of robot paths increases by implementing an offline programming tool. The other application of virtual which was investigated in this thesis was virtual commissioning. Virtual commissioning is a method used to test and validate robot and control (PLC) programs in a virtual environment before implementing in the real plant. The software in the loop was selected as an approach for building the virtual commissioning model after presenting the different existed approaches. The result of the model showed great benefits of virtual commissioning not just in reducing the time of traditional of the new built line but even when replacing parts in an already existing manufacturing. Virtual commissioning is still new in the manufacturing, therefore, there are many issues which should be solved to get an optimal model that imitates perfectly the real plant.

This work is being carried out in collaboration with the TYRI Light industry to improve their existing assembly line, which currently has 8 stations. Out of these, three stations are operated by humans, and the rest are automated. The main focus of this project is on the first station, which is currently operated manually. While another master’s student from University West focuses on improving an automated second station and enhancing the fourth station, which is operated by a human worker. The seventh station, which operated manually, requires a human worker because of its complexity and flexibility of human.At this first station, there are two tasks involved. The first task is picking up components and placing them on an empty fixture, while the second task is picking up the finished product from the fixture and placing it on the palletizing process. Currently, it takes 19 seconds for a human worker to complete these tasks before sending the fixture to the next station.The goal of this project is to investigate whether this first station can be improved by using a robot instead. If a robot can perform these two tasks faster than 19 seconds, enhance workplace safety, and handle both tasks effectively without a need for more robots or external assistants. To achieve this, a discussion was made with the TYRI supervisor to select a collaborative robot for this task. To validate the work, ABB RobotStudio simulation software is used. The collaborative robot should be equipped with a specially designed gripper, which is suitable for the collaborative mode. Furthermore, in order to make the gripper suitable for TYRI’s components measurements, it was customized by a mechanism feature, which RobotStudio has.With the help of literature studies and various methods, the successful implementation of the robot was achieved. The robot was integrated outside the assembly line and was able to handle both tasks at the first station. Which indeed improved workplace safety. However, the robot took a total of 30.8 seconds to complete the tasks, which is longer than the 19 seconds goal was aimed for. As a result, the robot didn’t meet the desired cycle time required in this case

Food Industry Automation is one of the sectors that took an advantage of industry fourth revolution, where food industry robots start to play a significant role in planting, transporting, processing, and even preparing various dishes. Pizza making is one of the food sectors that is entirely done by food experts and pizza chefs in small and medium business pizzerias, and even though pizza factories have mass production of pizza, they still don’t produce the required freshness and good taste pizza like the one you find it in pizza restaurants. The project studies food industry automation and then designs a conceptual model for robotizing pizza making process, providing a fully automated solution for medium and small pizza-making bakeries. Demonstrating what technologies are required to implement the idea and how to overcome challenges such as quality assurance and preserving hygiene standards adds to that, showing technical limitations and opportunities. The project study is based on theoretical and experimental aspects using an application development kit, Arduino IDE, and simulating the whole process with ABB robot studio simulation. 

In this work, a ROS (Robot Operating System) version of Open RatSLAM, [1] [2], was tested with Viso2 [3] as an alternative monocular odometer. A land based rover [4] was used to perform data acquisition and a remote control tool was developed to facilitate this procedure, implemented as ROS nodes on both Ubuntu 16.04 and on Android 7.0.An additional requirement that comes from using Viso2 is the need for camera information together with the image stream, which might require camera calibration. A ROS node to manually add this camera information was made as well as a node to change the generated odometry message from Viso2 to a form that RatSLAM uses. The implemented odometer uses feature tracking to estimate motion, which is fundamentally different to matching intensity profiles which the original method does and can hence be used when different properties of the visual odometry function is desired. From experiments, it was seen that the feature tracking method from Viso2 generated amore robust motion estimate in terms of real world scale and it was also able to better handle environments of varying illumination or that contains large continuous surfaces of the same colour. However, the feature tracking may give slight variations in the generated data upon successive runs due to the random selection of features to track. Since the structure of RatSLAM gives the system ability to make loop closures even with large differences in position, an alternative odometry does not necessarily give a significant improvement in performance of the system in environments that the original system operates well in. Even though both algorithms show difficulty with estimating fast rotations, especially when the camera view contains areas with few features, the performance improvement in Viso2 together with its ability to better maintain the real-world scale motivates its usefulness. The source code, as well as instructions for installation and usage is public

In this work, a ROS (Robot Operating System) version of OpenRatSLAM, [1] [2], was tested with Viso2 [3] as an alternative monocular odometer. A land based rover [4] was used to perform data acquisition and a remote control tool was developed to facilitate this procedure, implemented as ROS nodes on both Ubuntu 16.04 and on Android 7.0.  An additional requirement that comes from using Viso2 is the need for camera information together with the image stream, which might require camera calibration. A ROS node to manually add this camera information was made as well as a node to change the generated odometry message from Viso2 to a form that RatSLAM uses. The implemented odometer uses feature tracking to estimate motion, which is fundamentally different to matching intensity profiles which the original method does and can hence be used when different properties of the visual odometry function is desired. From experiments, it was seen that the feature tracking method from Viso2 generated a more robust motion estimate in terms of real world scale and it was also able to better handle environments of varying illumination or that contains large continuous surfaces of the same colour. However, the feature tracking may give slight variations in the generated data upon successive runs due to the random selection of features to track. Since the structure of RatSLAM gives the system ability to make loop closures even with large differences in position, an alternative odometry does not necessarily give a significant improvement in performance of the system in environments that the original system operates well in. Even though both algorithms show difficulty with estimating fast rotations, especially when the camera view contains areas with few features, the performance improvement in Viso2 together with its ability to better maintain the real-world scale motivates its usefulness.  The source code, as well as instructions for installation and usage is public.

Nowadays the globalization of the market has a direct impact to the companies. On one hand, gives them the opportunity to spread their business but on the other hand increases the competition and the efforts to succeed. This project is pursuing to investigate and design a system capable to take upon and optimize the production within functional shops by in-troducing automation. Although the system it would be a standard solution for the functional shops, a case study for GKN Aerospace Company with its specifications has been carried out. The task become trickier due to the aeronautic sector requires a lot of work and effort to produce their products. The items and materials are expensive as well as the processes and treatments thus the introduction of automation is quite complex. This case study has accom-plished a research through the concepts and options to introduce automation within this production process. Taking under consideration the costs and efforts that would demand the investment for automation, the project propose a new and innovative manufacturing system "Move & Play", suitable with the features and requirements GKN Aerospace Company may need.The system proposed has been designed regarding to features such as movability, modularity and flexibility. These features have been considered essentials in order to make the system worth it and cost effective. The job-shop approach that the company use in its production contemplate the continuous change of products and introduction of new ones. Hence, the idea is to use the "Move & Play" system when it is required along the production. As a consequence, this system will create a "local product flow" along that point in the production that will organize the rest of the production or "global flow". Moreover it is expected to apply the concept of "plug-in" the processes that the system need at each time and therefore give flexibility to the system. With that purposes, the work has studied three different designs of the "Move & Play" system. Each design have considered different aspects and disposition. These designs have been compared according with criteria such as geometrical, the capacity and investment needed and the cycle time spent. Finally a conclusion on which design is the most suitable has been stated as well as a future work to continue with the work has been proposed.

This project is being implemented in cooperation with TYRI Light Company, the company is serving many industrial sectors such as construction, mining, forestry, and transportation. The goal of this project is to provide a simulation of the automatic assembly station for as part of assembly line which assemble a product named 0909 compact lighting. Currently, with semi-automated production line, the average of production is 800 lights per a day (8 work hours), and three glue robots are utilized but componentsare manually assembled and fed. The emphasis will be on automating assembly and feeding as well as developing a mechanical design and a new glue dispenser. the validation and testing of all concepts will be in simulation environment. The fundamental idea behind this approach is flexibility, where just a little adjustment is needed whenever a new variable is added to the assembly line. After validating all the concepts, it turned out that the most suitable robot is articulated IBR 140T, and the mechanical tool with the distribution tool gave promising results, and the feeding system needs a lot of experiments, and it turned out that it is difficult to reduce the cycle time of a full assembly to one minute. But in order to benefit from the proposed solutions, the solution must be generalized to include the rest of the stations, allowing to reduce labour.

The degree project focused on automating industrial processes in heavy truck manufacturing using machine learning (ML) and computer vision. Specifically, it addressed the challenge at the Volvo Trucks plant in Tuve, Gothenburg, of accurately measuring the height of finished trucks. Manual measurement methods that were in use were prone to errors and inefficiencies, prompting the need for an automated solution implemented through the Volvo Vision System (VVS).

The project developed and evaluated a proof-of-concept (POC) system that could be integrated into VVS for automated truck height measurements. The system incorporated camera calibration, ML models, and computer vision techniques. Testing on various truck models and configurations demonstrated that the system provided reliable detection of trucks cabs, variations of addons that were random and variable in geometry and perform measurements within an acceptable error margin 90% of the time, with a mean deviation of 2.25 cm and a standard deviation of 2.36 cm. The automated system significantly outperformed manual methods, reducing the measurement time per truck from 60 seconds to 120 milliseconds.

The study concluded that the enhanced VVS could reliably measure truck heights within the required industrial error margins. The automation of the measurement process not only increased accuracy but also improved efficiency, aligning with Volvo’s goal of reducing man-ual interventions.

The project highlighted the potential of integrating AI and computer vision in industrial applications, paving the way for future enhancements and broader applications in similar settings. Future research could focus on expanding the system's capabilities and adapting it to other industrial environments.

Company Purtech in Dals-Ed manufactures molded polyurethane (PUR). Examples of products that include polyurethane are car seats. Robots are used to fill the molds with PUR and to apply the release agent (wax) in the empty molds.

Turning from online programming into a graphical offline programming of release agent spraying robots is going to simplify the process by:

1. Applying less of release agent to avoid polluting environment, to produce an easier removal of moulds, for the sake of homogeneous moulds and for economical saving in the cost of release agent
2. Adaption of spraying paths to variation in production speed.
3. Programming of complex spraying trajectories to deal with sharp geometrical subsurface
4. Decreasing onsite programming time (when program a new workpiece or modify an old one); so that robots would be free for production.

While turning into offline programming brought the challenges of:

1. Impact of variation in the production speed
2. Lack of 3D models of workcell’s equipments
3. Robot joint configuration when paths and robtargets are in move.
4. Physical Joint limits, Singularities & Reach limits
5. Collisions within the cell space.

At the end, the following objectives are successfully met:

1. Adaption of spraying programs to variation in production speed by developing and embedding a method in those programs.
2. Graphical offline generation of spraying trajectories and optimization of those trajectories to the Purtech condition of spraying allowed time for each carrier.
3. Simulation of release agent spraying process; and producing of a well structured RAPID program that reflect the simulated process.

This thesis project introduces machine vision to automate the testing and verification of watch hands and its digital display functionalities. Conventionally, the verification process in this watch manufacturer involved manual inspection by operators, a method that is time consuming and involves many manual phases to get done. This project introduces a level ofautomation to increase testing efficiency by introducing one standalone environment and testing capacity by increasing production capability and speed.

This project included a development of a prototype testing environment built from 3D printed components, controlled by a RaspberryPi 4 model B processor, and utilizing its default camera module for achieving the machine vision part necessary for automation. The project's aim is: first, to verify the testing of watch hands' functionality, and second, to ensure the OLED display operates as intended, utilizing image processing techniques via machinevision

Rapporten som redovisas här är ett examensarbete med uppdrag att genomföra förebyggande underhåll i kross- och separationsavdelningen på en batteriåtervinning fabrik. Fabriken återvinner litiumjonbatterier som används i elbilar. I kross- och separationsavdelningen är det mycket buller och damm som en följd av skakande och krossande processerna i den mekaniska separationen. Det är tunga maskiner som genom olika operationer sliter isär metallkomponenterna i batteriet till mindre delar för att slutligen mala dem och separera beståndsdelarna i materialet åt med avseende på deras magnetiska egenskaper. Förutsättningar i kross och separationsavdelningen medför ohälsosam miljö för människor, vilket leder till att avdelningen måste styras automatiskt utan att medarbetare behöver befinna sig i avdelningen. Det föreslagna förebyggande underhållssystemet som behandlas i rapporten stöds av teknik från Industri 4.0. Industri 4.0 är benämningen på en rad moderna teknologier och koncept som används inom automation. Maskin-till-Maskin kommunikation och Internet of things(IoT) är integrerade för högre grad av automation. Förebyggande underhållsarbetet som presenteras baseras på en förstudie analys som nischar sig inom förebyggande underhåll och industri 4.0. Förebyggande Underhållsarbetet som redovisas i denna rapport kommer inkludera ett signalflödesschema som ska visualisera kommunikationen mellan enheterna i avdelningen,en schematisk design för övervakningssystem, ventilationen och ett underhållsschema för maskinerna i avdelningen

A collaborative Plug&Produce system is a flexible, modular, and reconfigurable system that helps cope with fluctuating market demands. The idea is to be able to reconfigure the production line in minutes instead of hours or days. This reconfiguration is made by adding or removing a module to achieve different manufacturing goals. Guaranteeing the safety of this system is a crucial question in achieving the desired flexibility, as any change in the system requires a new risk assessment. The risk assessment process is still conducted manually, and it is a time-consuming process. This thesis work is a part of several studies carried out at University West to create an automated method to conduct the risk assessment in a way compatible with the collaborative Plug&Produce system. In University West studies, the Plug&Produce system is controlled by a Configurable Multi-Agent System (C-MAS). Particularly, in this thesis work, a concept is developed to automatically generate the hazardous zones of the collaborative Plug&Produce system to facilitate the process of hazards identification during the risk assessment process after every reconfiguration. The needed data to generate the hazardous zones are related to two systems. The first system is the collaborative robot system, where “Speed and separation monitoring” is the collaborative method used in this work. A protective separation distance (PSD) will be maintained between the robot and the operator to avoid collisions. The second system is the Plug&Produce system controlled by C-MAS, where the system consists of parts and resources. The parts have goals to be implemented, and the resources have the skills to implement the parts' goals. The resources have different variables that help describe the resources and ensure the compatibility between the resources to perform a particular sequence of skills called a process plan. These resource variables are saved in a database and help generate the hazardous zones. Additional data related to the Plug&Produce system controlled by C-MAS is also needed to generate the zones, such as the table of the identified hazards and data about the hazardous zones of the adjacent modules.

The hazardous zones are represented in this work as bounding volumes describe the volumetric occupancy of the robot’s moving link. The cuboids are chosen as bounding volumes to represent the hazardous zones.The developed concept is tested on a cell to produce a wooden walls simulation created in Robot Studio. Two process plans that perform the same goals are used in this test, and the results have been compared and discussed to validate the proposed concept.

Parker Hannifin Manufacturing Sweden AB is an international company with manufactory sites all around the world. The site in Trollhättan is part of the Pump & Motor Division Europe. The Factory has worked a lot with automating the manufacturing section. Many robots are implemented in the factory to produce the simple parts which need to be assembled to interduce the last product. Parker are now interested in starting the process of automating the assembly stations for the 3 products they produce in this site. The three products are Truck pumps, fixed and variable displacement pumps (LBP), variable hydraulic motors (VM) and fixed hydraulic motors (FM). Each one of the three types has different sizes and different series.

This work focuses on analysing the current system which consists of seven assembly station, 101, 102, 201, 202, 203, 301 and 302. After analysing the current system, a work on finding the potentials for moving from manual to semi-or automated stations for each station is presented. Seven potentials were found. The first potential which handle a bout a central pressing robot cell is a common potential for all the stations, and the other potentials are for each station separately. Testing and validation for all the potentials where the automation can be implemented and started with, the cycle time for the operations and the productivity, the workspace and material handling in addition to the level of the ergonomics for these stations.

The complexity of these solutions, the advantages and disadvantages and a safety analyses are also presented. And how to optimize the current system for reaching a high level of automation is discussed.

The Internet of Things (IoT) concept is getting attention recently, and the number of connected devices is rising rapidly, climbing above one billion. Combining both IoT and Robotics ends up with the Internet of Robotic Things (IoRT). This paper aims to explore the ethical standards of IoT and Robotics to build up the ethical margins and structures of IoRT in the healthcare sector. Thus, an ethical pyramid consists of security, privacy, and safety in such systems; this pyramid helps to understand the current ethical principles and how to adjust and translate them into protocols suite IoRT in the medical sector. These protocols are formulated in the form of responsibilities and rights of IoRT subdivision which is Internet of Medical Robotic Things (IoMRT

The manufacturing industry and associated systems are being developed in an increasing manner to cover the market needs, where the manufacturing companies are continuously racing and competing to achieve high productivity rate and better production quality. In this thesis work, an advanced method for motion control of industrial robots has been investigated and implemented. This method is based on motion tracking of a 3DSpaceMouse, which was used to perform movements by the operator. The benefits and disadvantages of this method were discussed in this thesis work. It mainly showed a high accuracy in response to the motion applied by the 3DSpaceMouse, and a great stability regarding the programming environment that was used to build it. The movements applied by the 3DSpaceMouse were successfully captured and stored in variables in the programming platform. The capturing and storing process was successfully created as a package and prepared to be exported for usage by other software. Complete simulation was performed for an industrial robot, and successful communications among the various hardware and software components of this solution were accomplished. This has formed a complete integrated solution that has also included a user-friendly HRI. This HRI made it easy and simple to track the motion control processes and establish connections with the robot controller. Thus, it can be considered a feasible solution for motion control of industrial robots, which can be used by the manufacturing companies. Several tests and verification processes were carried out to obtain accepted results and to succeed in implementing a working model. Some errors and unexpected events have appeared during the work, which required handling in order to achieve a working integrated system.

Multi-agent systems (MAS) have gained a lot of attention among researchers during the last decade, they have also been widely used within a variety of industrial fields including robotics, automation, and automotive industries. A multi-agent architecture could be used for designing systems that perform complex tasks which cannot be handled in a monolithic way, such architecture ensures that different participating agents (hardware or software) are fully decoupled from each other which makes it easier to design, program, and deploy them separately, this also makes the complete system fully dynamic and capable of handling later structural changes like adding, removing or changing the behavior of any of the participating agents. As the agents are not aware of each other’s existence in the system, The gained advantages when using such a multi-agent architecture come with difficulties related to the interactions between the system agents, the agents need to register themselves in the system, share information about their skills, abilities and provided services with all other agents inthe system, only this way their services could be made use of by the other agents. These difficulties make configuring, programming, maintaining, and modifying multi-agent systems a hard and time-consuming task to be accomplished even when using advanced development frameworks which natively support the ability to configure multi-agent systems such as JAVA Agent DEvelopment Framework (JADE) and Robot Operating System (ROS).

In this master thesis, a MAS text-based modeling language is proposed. Moreover, a complete tool suite for modeling, and automatically programming ROS-based multi-agent systems is presented. The defined MAS modeling language and its corresponding tool suite aim to make the process of designing, programming, compiling, and installing the software of multi-agent systems easier and more budget and time efficient for companies and researchers. A standard and relatively complex example of a multi-agent system from the automotive industry was presented and modeled using the proposed MAS modeling language. Later, the implemented tool suite was used to generate code out of the model, compile it, install it and run it in a Linux environment, the generated code was analyzed using a static code analyzing tool and different code metric values were registered. The executed final binary format of the generated software was also analyzed during run-time using ROS graphical system monitoring tools. Future system maintenance and adaptations were simulated by manually modifying the initially created automotive system model and introducing system updates to it, the software was regenerated, and the static code and software run-time analysis were reperformed, results from both analysis reports were compared to the corresponding analysis reports which were created using the previous version of the system model. The results have shown that the proposed modeling language can be used for modeling large and complex multi-agent systems from a real-world example. The implemented toolsuite works as expected, it generates ROS-based software for multi-agent systems which is compilable, installable, and executable. During run-time, the software also behaves according to how the system was modeled. Finally, the registered statistical and code metrics measurements have proven that using such a modeling approach for automatically programming multi-agent systems reduces the development and maintenance efforts of these systems significantly

This study investigates the integration of Agile approaches with hightech advancements such as digitalisation, automation, interconnected systems, and IoT within the automotive industry.

The primary aim is to provide a comprehensive framework for Agile approaches tailored to the advanced automotive industry. It addresses the growing need for adaptability in development processes to meet market demands and rapid technological overevolving as vehicles transform into computers on wheels.

The study begins with a thorough literature review of both Agile approaches and emerging industrial technologies. This theoretical foundation identifies gaps in existing research and underscores the necessity for a framework to integrate agile approaches with high techniques.

Empirical data were collected by a qualitative case study, through interviews with professionals from large, global automotive companies, including Volvo Cars and Volvo Construction Equipment (VCE). These interviews provided realworld insights into Agile approaches' opportunities, challenges, and tailoring to meet industryspecific needs.

A framework for agile implementation in the automotive industry is constructed and comprised of an extensive literature review and profound insights from the results of the content analysis of the interviews. This framework emphasises cultivating an Agile mindset, leveraging modern technologies, fostering customer collaboration, streamlining documentation processes, embracing changes promptly, improving planning methodologies, increasing flexibility, and acquiring relevant skills.

Key findings reveal that process development methods must evolve alongside advancing technologies. Industries must adapt by tailoring Agile approaches and incorporating hybrid approaches, as reverting to previous methods is not a choice.

This research contributes to both theory and practice by providing a comprehensive framework for Agile approaches in the automotive industry. It bridges the gap between hightech solutions and Agile approaches, offering actionable insights for enhancing the development process, operation and operation, and competitiveness. Despite its contributions, the study acknowledges limitations, such as focusing on large automotive companies in Sweden and relying on just only a qualitative approach. These factors may impact the generalizability of the findings.

Future research should expand the scope, incorporate mixed methods, and validate the framework in diverse industrial settings.

Design of experiment (DOE) is a statistical method for testing effects of input factors into a process based on its responses or outputs. Since the influence of these factors and their interactions are studied from the process outputs, then quality of these outputs or the measurements play a significant role in a correct statistical conclusion about the significance of factors and their interactions. Linear regression is a method, which can be applied for the DOE purpose, the parameters of such a regression model are estimated by the ordinary least-squares (OLS) method. This method is sensitive to the presence of any blunder in measurements, meaning that blunders significantly affect the result of a regression using OLS method. This research aims to perform a robustness analysis for some full factorial DOEs by different robust estimators as well as the Taguchi methodology. A full factorial DOE with three factors at three levels, two replicants, and three replicants are performed is studied. Taguchi's approach is conducted by computing the signal-to-noise ratio (S/N) from three replicants, where the lower noise factor means the stronger signal. Robust estimators of Andrews, Cauchy, Fair, Huber, Logistic, Talwar, and Welsch are applied to the DOE in different setups and adding different types and percentages of blunders or gross errors to the data to assess the success rate of each. Number and size of the blunders in the measurements are two important factors influencing the success rate of a robust estimator. For evaluation, our measurements are infected by blunders up to different percentages of data. Our study showed the Talwar robust estimator is the best amongst the rest of estimators and resists well against up to 80% of presence of blunders. Consequently, the use of this estimator instated of the OLS is recommended for DOE purposes. The comparison between Taguchi’s method and robust estimators showed that blunders affect the signal-to-noise ratio as the signal is significantly changed by them, whilst robust estimators suppress the blunders well and the same conclusion as that with the OLS with no blunder can be drawn from them.

The use of wood, such as Cross Laminated Timber (CLT), has increased over the last years in the construction of buildings. CLT is primarily considered a structural component inside walls, floors, and ceilings. But, when the design specifies an exposed surface of the CLT, where the wood surface will be visible in the finished building, there are requirements on the visual appearance of the CLT. To fulfil the requirements visual inspections of wood defects of the surface are conducted, before the CLT-panel is shipped to the construction site. Human inspectors are not consistent evaluators of products, and their accuracy depends on factors such as expertise, motivation, and fatigue. This study proposes a solution to automate the inspection, where captured images are analysed by an object detection model, based on deep learning. The model Faster R-CNN has proven to be successful in medical image processing and for object detection. A challenge with deep learning models, such as Faster R-CNN, is the number of test images needed to train the model. Also, there are a limited number of suitable test images of wood defects available. By using a pre-trained Faster R-CNN model and adapt the model's skill to detect wood defects, instead of the model's intended objects, the model can be trained with a relatively small number of test images. This study relies on images of planks, similar to the ones used in CLT, and the number of test images in the study is very limited. To improve the results, the model needs to be trained with more test images of CLT-panels. Nevertheless, the model can locate the defects in the CLT-panel and identify the type of defect to a high degree. The results indicate that the proposed model can be used to automate the inspection of wood defects on the CLT surface.

The main goal of the thesis is to create a way to interact with a ABB robot by using a Kinect.The focus will be put on making the robot mimicking the user's arm movement. But otherway like voice control will be investigated.Two main part of the work are developed in this report. The first one is the software aroundthe Kinect to get the body data. The software developed provides the user a visual feedbackon what the Kinect see and allows him to choose what mode he wants to use. Two mainmode are implemented. A mimicking one which makes the robot mimicking the chosen armand a moving object mode which tell the robot to go to the user's position and set the toolas the user hand state.When all the data are gathered, they are sent to the robot using an IP socket and predefinedmessage to give him the desired order.

This thesis focuses on the creation of a shape detection algorithm designed to work with 3D point cloud data, which is essential for many industrial automation tasks. The algorithm was developed using traditional image processing techniques and was tested on a variety of shapes, ranging from simple forms like spheres and cones to more complex objects such as dice or cubes. The goal was to create an efficient and accurate algorithm that can detect these shapes in real-time, which is crucial in industries where speed and precision are important.To ensure the captured data was accurate, the Photoneo 3D scanner was carefully calibrated. This scanner was used to generate point cloud data, which is essentially a digital representation of the surfaces of objects. The data was then processed by the algorithm, which identifies and refines the shapes it detects. The results of the testing showed that the algorithm is particularly good at detecting simple geometric shapes like spheres, where it performed with high accuracy. However, when it came to more complex shapes like dice, the accuracy dropped slightly. This drop was more noticeable when multiple objects were present in the data, as the algorithm struggled to distinguish between closely positioned objects.In terms of performance, the algorithm was efficient, especially when handling simple shapes. It could process the data quickly enough to be used in real-time applications. However, more complex shapes took longer to process, and the algorithm required more memory to handle these cases.

The study also considered how the algorithm compares to other methods, such as traditional image processing and deep learning approaches. While deep learning techniques tend to offer higher accuracy, they also require much more computational power and memory, making them less practical for real-time applications in some cases. On the other hand, traditional methods are faster and require less memory but may not handle complex shapes as effectively. The algorithm developed in this thesis strikes a balance between these approaches, offering good performance with moderate computational demands.Additionally, the study aimed to incorporate the Volume of Interest (VOI) algorithm into the detection process. However, it was found that the VOI algorithm did not perform as well as expected, particularly due to its sensitivity to noise in the data. As a result, the outcomes from the VOI algorithm were not included in the final analysis, indicating an area for future improvement. The thesis concludes that the developed algorithm is effective for most scenarios but could benefit from further improvements, such as incorporating deep learning techniques to handle more complex environments. This work has practical implications for industries that require accurate and fast shape detection, such as in quality control or automated assembly lines. Future research could focus on optimizing the algorithm further, testing it with more varied and complex shapes, and exploring how it can be integrated with other technologies for even better performance.

This project aims to present a condition-based maintenance algorithm for common motor faults on three-phase induction motors in the mining industry through the use of the E300 overload relay. To fulfil this aim establishing what motor faults are common and which ones are suitable for detection based on electrical signal monitoring was explored. Through investigative research no distinction between common motor faults in general and those in the mining industry was discovered. A number of electrical signals candidates were chosen based on limited available research and consulting with experts in the field. To test these signals a number of phase to phase motor faults were physically simulated on a test rig. The resulting trend of the signals were saved and comparatively analyzed with a regular mechanical motor overload to find deviations that would indicate the fault has occurred.

The most useful monitored signal was the current imbalance and the line current measurements. Through monitoring these signals it was concluded that the one phase-pair fault and the two phase-pair fault could be discovered by monitoring real-time values. The three phase-pair fault could not be discovered by monitoring the real-time values, but can be found trough historical trends of the line currents and current imbalance. 

A condition-based maintenance algorithm was proposed utilizing the monitored electrical signals to create trends that can be tied to alarms. A general motor health index that would use the monitored signals to easily convey the health of the motor is proposed. A faceplate design for the Panelview 5510 was designed showing the trends of the various electrical signals that would be used by an operator in the field.

The background for this thesis work emanates from a desire to drive the production of the VBG Truck Equipment factory in Vänersborg a step further to catch up with the Industry 4.0 standards. Discrete Event Simulation (DES) is a widely adopted tool within industries to model real-world systems to improve their processes. The concern of bottlenecks is a vital problem in optimizing and improving the efficiency of production processes. Finding and investigating bottlenecks is one of the primary considerations of all manufacturing industries.The company aims to enhance its logistics by using automation in its processes. This work focusses on the Paint shop part of the VBG factory. Thus, the VBG would like to analyse the methods to reduce the bottlenecks and the waiting time in the Paint shop production process. There were no simulation models available to analyse the existing paint shop pro duction flow; therefore, a Discrete Event Simulation (DES) model of the VBG Paint shop was built to investigate the bottlenecks and improvement methods.

In this thesis work, the DES model of the existing Paint shop that uses the Last in First Out (LIFO) method was investigated to find the throughput and lead time for the products. In comparison, a First in First Out (FIFO) method was used to get a balanced production flow, and its effects were studied. The company currently uses two shifts in their Paint shop, so the alternate arrangement of shifts using the same number of operators was analysed. The analysis of three shifts with the same number of operators gave more efficiency than the existing one, and three shifts with an increasing number of operators were also compared. The shift arrangements were compared with both the LIFO and FIFO methods, and FIFO provided more throughput and balanced production flow.

The main bottleneck was identified in the forklift transport waiting time, so an alternate layout using a power and free conveyor system was developed. This alternate arrangement reduced the bottlenecks and produced an increased output. The alternate layout was also compared with LIFO and FIFO methods; FIFO was the most efficient one. The shift patterns were also analysed, and the three shifts with a 50 percent increase in staff can reduce the stock build-up during the start of the first shift. As of this study, it was identified that the alternate model with FIFO is the most suitable solution for the company, and in the coming years, they intend to implement that. From the experimental results obtained, DES can be chosen as a method to analyse the problems associated with limiting the capacity of production. The DES modelling can be extended to the other processes in the factory and can be used for improving logistics and inventory management. The company production flow can be further advanced by evaluating the changes that can be implemented in the factory with real-time data.

The work explores the possibility to increase the automation along a production line by introducing robots without reducing the safety of the operator. The introduction of a robot to a workstation often demands a redesign of the workstation and traditionally the introduction of physical safety solutions that can limit the access to the work area and object on the production line. This work aims to find a general solution that can be used not only in the construction industry, but also in other types of industries to allow for an increased Human and Robot Interaction (HRI) without physical safety solution. A concept solution of a dynamic and flexible robot cell is presented to allow for HRI based on safety zones in a shared work environment. The concepts are based on one robot and the usage of a 3D camera system allowing for the design of virtual safety zones, used to control the HRI. When an operator approaches the robots work area and triggers a safety zone the robot stops its work and moves away from the operator. Based on the safety requirements and triggered zones the robot will continue to work in a new area or wait until the operator leaves the work area and then continue with the interrupted work task. This will allow the operator and the robot to work together, where the operator location controls the robots workspace. Testing and validation of the presented concept showed that the wanted functionality could be obtained. It also showed limitations to the equipment and the system used during tests and raised additional aspects of the safety for HRI. Of the detected limitations the most crucial when looking at up-time for the production line, is the camera system need of a relatively dust free environment, good and constant lighting. For the safety of the system the limitation lies in the size and placing of the safety zones in combination with the disturbance from  surrounding equipment. The presented concept has proven to work, and can be applied not only for the construction industry but for all industries with manufacturing alongside production lines with large components.

In this paper, early work on how to implement flexible safety zones is presented. In the case study an industrial robot cell emulates the environment at a wall construction site, with a robot performing nailing routines. Tests are performed with humans entering the safety zones of a SafetyEye system. The zone violation is detected, and new warning zones initiated. The robot retracts but continues its work tasks with reduced speed and within a safe distance of the human operator. Interaction is achieved through simultaneous work on the same work piece and the warning zones can be initiated and adjusted in a flexible way.

Flexible human-robot industrial coproduction will be important in many small and middle-sized companies in the future. One of the major challenges in a flexible robot cell is how to transfer information between the human and the robot with help of existing and safety approved equipment. In this paper a case study will be presented where the first half focus on data transfer to the robot communicating the human's position and movements forcing the robot to respond to the triggers. The second half focuses on how to visualize information about the settings and assembly order to the human. The outcome was successful and flexible, efficient coproduction could be achieved but also a number of new challenges were found.

The use of robot has a major part in all kinds of industry (food, textile, automobile, agriculture, space and so on). With the increase of robot in industry, robot programming is also becoming more and more important, and more complex, therefor not accessible to anyone. The common way to program robot is by using programming languages, to communicate with the robot and its different element, and simulation, to observe the robot behaviour in a simulated environment before using a tangible one. Programming languages can be tedious, especially robot programming, which means that it can only be done by someone with the proper knowledge and education, making robot programming almost impossible to understand for non-robot programmers. To tackle this issue, a simpler programming method has been implemented during this thesis work, by combining machine vision and sensor technology. The idea is to track, using machine vision movement of a human arm, and use sensors, in this case a gyroscope, to register the orientation of the arm. The conclusion of this work shows us that the method investigated and developed, could allow a user, to manipulate an industrial robot, without knowing anything about robot or robot programming. Although, this method can be enhanced, by recording the movement done by the user, so that it can reproduce it for different work, by using an accelerometer tomove the robot at the same speed as the human hand. 

The exponential increase in the use of Electric vehicles has a potential remark that there would be a massive count of EV battery packs reaching end of life. Such products being hazardous to environment, it is clearly proven to have processes to reuse or recycle the same. Northvolt AB leading Li-ion battery manufacturer in Europe is striving towards developing green battery technology. Besides being a leading industry to produce batteries, they have realised the need of recycling EV batteries at an early stage and developed a pilot-scale recycling plant now to close the loop of battery manufacturing. In pilot plant, the primary step to handle EOL EV batteries is to dismantle them in an automated manner. The present thesis work contributes towards developing the automated dismantling process to be more flexible and reliable in handling heterogeneous battery packs arriving from various car manufacturers. Out of many research works being carried out in the pilot setup, the current thesis work is focusing towards two key areas of product digitalization to obtain a digital model of real-world object and extracting data from a digital model to provide a data input to the control systems of automated disassembling process. This clearly eliminates the methods of lower-level programming of automated systems and saves time and resources involved in re-programming and increase flexibility in the process.

The inspection of production parts is a difficult, tedious and exhausting task, usually performed by human experts. Especially when it comes to small features, a manual inspectioncan be insufficient and susceptible to errors. To aid the quality control of transparent injection molded test stripe components and validate the color accuracy of installed reaction papers, a machine vision system is designed and build for the automatic inspection of the givenparts. This work applies a defect detection and classification approach which makes use ofthe difference in appearance of certain defect types depending on the way they are illuminated. In a following steps filters and other image processing methods are used to separatethe defect from the rest of the image. For each defect type, a specific lighting method andprocessing approach is chosen, which allows it to filter out all other defects. This approachis complemented by a convolutional neural network, trained for the classification of the teststripe components into ok and nok parts. For the task of color validation, a color calibrationalgorithm using a color chart is implemented and used as a basis for a referencing approach.In this approach the acquired color values of the sample are compared to a reference table,which allows it to translate them into pH or nitride values.The resulting system is capable of classifying the injection molded components with anaccuracy of over 98%, while classifying defects into dust, scratches, surface defects, distortions and inclusions. Working with the Lab color space, the implemented color validationachieves a root-mean square error of 6.35 for L, 2.77 for a and 4.36 for b, which translatesto an error of 0.2 when tested on a pH sample.

Manufacturing and automation lines are getting expanded to include Digital Twins as stepstowards Industry 4.0. In this regard, apart from the issues that Digital Twin designers commonly experience during the process, implementing the virtual model on the real system while transferring the codes and data can be challenging as well. Using Digital Twin as part of the future laboratory facility of University West for the automation line at Production Technology Centre (PTC), it first needs to be tested on the real line to discover possible problems that future students might face. In this master thesis, these errors and faults are investigated to help the automation line staff and students avoid them by possible solutions.The assembly routine in the Digital Twin of the automation line was tested with an HMI using the OPCUA platform both in manual mode and automatic mode. The automation line was tested in manual mode by changing the function blocks which work with the real PLC. Moreover, the students can use the conversion table to understand the difference between the Digital Twin and its real entity function blocks. However, it can take much time for students to make their HMI and PLC code compatible with the automation line at PTC. As a result, suggestions are made, and attention to some minor details of the test procedure are emphasized. Although, due to the COVID-19 pandemic, and packed schedule of PTC staff, the automation line was not completely assembled and for safety reasons the automatic mode of the tests was withdrawn.

Friction Stir Welding is a solid state join process in which the heat need is produced by the friction of a tool on the work pieces. The tool rotates on the surface and also applies a force against it to create the needed friction. This technology does not use any consumable and does not produce any gas, which makes it cost effective, save for the operators and clean.

The equipment used for this technology has been changing within the years. Milling machines were adapted to it at the beginning. After, specific machines for friction stir welding were developed, providing higher down force and control systems. Nevertheless, the lack of flexibility of machines made industrial robots became an alternative. Robots have limitations in terms of stiffness. But, they can be used to weld complex geometries and they can be used in large scale applications due to their flexibility and low cost.

This project studies the process from the robot behaviour perspective. The main objective is to achieve a better understanding of the process to contribute to the continuous development of robotic FSW. An acquisition system was developed to monitor different signals from the robot such as, forces, torque, path deviation, and vibrations. The data acquisition system developed and mounted at PTC was tested and its limitations were examined. The developed system has been used for conducting different experiments: The differences between the robot system and a system based on a FSW machine situated at TWI were studied; and, in addition, the behaviour of the robot under different welding parameters was investigated.

The Sawmill industry plays a crucial role in the forestry sector, providing the raw material needed for various industries such as construction, furniture, and paper production. Efficient production in sawmills is essential to meet the increasing demand for lumber and to remain competitive in the market. In this context, Discrete Event Simulation (DES) is a powerful tool for improving production efficiency by identifying bottlenecks, reducing lead times, and optimizing inventory levels.

This thesis project aims to develop a simulation model using Discrete Event Simulation (DES) to improve the production flow in a Sawmill. The simulation model allows the company to test different scenarios, such as changes in production schedules, machine configurations, and inventory management, without disrupting the actual production. By answering two investigative questions, the report provides insights into the level of information needed for the simulation and the limits of the simulation model's capability to detect flaws in the production process.

Accurate data is essential for building a reliable simulation model. Therefore, the report emphasizes the importance of collecting data on various aspects of the Sawmill, such as machine utilization rates, machine capacities, processing times, and inventory levels. The report also highlights the need for simplifications and assumptions to create an optimal model, as collecting data on every detail may not always be feasible. 

The ExtendSim software is used to build the Discrete Event Simulation (DES) in this report. This software allows the modeler to create an accurate representation of the Sawmill's production flow and analyze the impact of different scenarios on the production process. The simulation model can help the company make informed decisions by identifying bottlenecks, reducing lead times, and optimizing inventory levels.

In conclusion, this thesis project provides valuable insights into the use of Discrete Event Simulation (DES) to improve the efficiency of the Sawmill process. The simulation model developed in this report can help the company test different scenarios, reduce lead times, optimize inventory levels, and make informed decisions. The report highlights the importance of accurate data, simplifications, and assumptions in building a reliable simulation model. By leveraging Discrete Event Simulation (DES) as a tool for optimization, the Sawmill industry can enhance its productivity and remain competitive in the market.

This paper presents a method for automating the generation, verification and deployment of robot programs used in prefabrication of walls for family houses. The making of robot programs is today performed manually by experts, i.e. implying high costs. This is a huge disadvantage since each wall can be unique. The work demonstrates, with implementation and testing, a method to automate the generation of robot programs for fabrication of walls made of wood. This includes the task of generating collision free paths, automatic verification of path performance and deploying to a real industrial robot with minimal human interaction. © 2018 The Authors. Published by Elsevier B.V.

The concept of Automation began with the rise of Industrial Revolution 3.0. Automation was then done in a partial manner with help of computers and programmable controls. With Industrial Revolution 4.0, world started focusing on automation. Industries working on automating their factories and car manufacturers towards driverless cars. Nowadays, robots are being used for public services and transportation. For navigation purpose, these robots arestill using Global Positioning System (GPS), which is highly advanced but with its own uncertainties. To bypass the use of GPS in industry and implementation in autonomous cars, one can use combination of machine vision and photogrammetry techniques. The machine vision aspect is to detect multiple objects having known coordinates in wall coordinate system and identify the position of these objects by recognising their class value in an image. First step was to calibrate the camera being used for the work. It was done by using MATLAB’s inbuilt Camera Calibrator App, from where we derive the focal length of the camera. Next step was to detect and classify the objects present in the environment. After detection, the centroid values of each object in the image plane were determined. For the photogrammetry aspect, one should know these objects’ position in wall coordinate system, and then using the concept of collinearity and triangulation, combining it with the centroid values gained from machine vision aspect, one estimates the position of camera from which the video and images are being taken. The result of this part indicates a method where the camera can detect multiple objects in the image going from minimum of four to maximum of five with their centroid values and respective classes plotted on the image or the frames of the video. As the distance between the image and the projection centre is just the focal length of the camera, the method utilises the knowledge of basic trigonometry to determine angles between the projection centre of the camera and object coordinates in image and the angles are observed to be same for projection centre and actual objects in the environment. From the calculated distances between the objects in real time and the projection centre, one estimates the coordinate of the projection centre using a nonlinear least squares method. All these calculated values combined with the data gathered from the machine vision aspect, are used to estimate the camera position in an environment with some uncertainty. This uncertainty is the error in calculating the position of the camera. The result shows the proof of conceptual model and can be implemented by industries to overcome the issues faced with the use of GPS. 

By conceptually comparing the two simulation methods, this thesis work investigates the limitations and possibilities of DES and combined model in plant simulation and visual components. Also look into which methods are best for teaching simulation and how they affect learning outcomes. A common benchmark was created to compare the DES and combined model, and a focus group session was held at University West to gather information from participants about the model's limitations and possibilities. Using an combined model instead of a DES model solves complex problems, while a hybrid approach allows existing models to be flexibly combined within a common large-scale context. According to the focus group, plant simulation is used to quickly build a conceptual model; however, one of the most concerning limitations of plant simulation is rarely close to real time application. Visual components, on the other hand, are the inverse of plant simulation; the models that are built arerelated to real-time applications, and the most significant disadvantage of visual componentsis time consuming

This paper presents a concept for reconfigurable and flexible robot systems. To reach a technology readiness level where solutions and results can be implemented in industry, the focus in this work is on systems with limited number of robots, and work scenarios which are reasonable complex but hard to automate using standard solutions.Four distinct areas have been identified as important within the concept and further studies: (i) human machine interaction, (ii) safety including collaboration, (iii) programming and deployment, and (iv) planning and scheduling. Feasibility studies have been made which addressed issues (ii) and (iii), in scenarios with collaboration between robot and human, or between two robots. For the chosen work scenario, manufacturing of structures in wood for family houses, challenges related to programming and safety was identified and possible solutions outlined.The concept and the studies indicate that feasible solutions can be found and designed given a reasonable consistent work processes and products. In this study, the processes are similar, nailing and screwing but different sizes may apply, the material is similar but variations may apply, and the construction is different of each product, but include the same type of operations at different locations. Our study confirm that human collaboration improves the ability to implement and use robots as it make it possible to move some operations to the human which otherwise would add to the complexity of the system. Furthermore, programming can also I general be simplified although methods for automatic programming has been tried out. But in some cases, the solution space is limited and the ability to move certain operations to a human simplifies the programming task. However, further work needs to be done in this area specifically related to safety issues for safe collaboration.

The ongoing trend towards manufacturing of customized products generates an increased demand on highly efficient work methods to manage product variants through flexible automation. Adopting robots for automation is not always feasible in low batch production. However, the combination of humans together with robots performing tasks in collaboration provides a complementary mix of skill and creativity of humans, and precision and strength of robots which support flexible production in small series down to one-off production. Through this, collaboration can be used with implications on reconfiguration and production. In this paper, the focus and study is on designing safety for efficient collaboration operator—robot in selected work task scenarios. The recently published ISO/TS 15066:2016 describing collaboration between operator and robot is in this context an important document for development and implementation of robotic systems designed for collaboration between operator and robot.

Collaborative robotic systems where human(s) and robot(s) cooperate in performing a common task is an attractive solution to introduce automation combined with high flexibility for tasks that have a high complexity and characterized by low volume or down to one-off. By introducing collaboration in robotics systems, the operator can complement with cognitive capacity and skill in order to gain in flexibility and agility in the task operation. This paper describes on-going work related to work on collaboration between operator and robot. User scenarios are outlined together with methods, software components and hardware to support collaboration, where some of these are under development. As the standards related to collaborative robotic systems are soon to be completed, it is expected that this type of semi-automatic systems will be important for flexible and agile automation of production which otherwise cannot be automated.

Automation is a broad topic with many difficulties. Without extensive knowledge of the subject, it’s easy to forget important aspects when a production line is to be automated. This degree project aims to assist companies with this process by constructing an assessment matrix that will from asking different questions generalize the intended process and give feedback on a suitable strategy for automating it. The assessment matrix is in this degree project tested with Ultramare AB, a Swedish producer of air filters, in one of their manual production lines. To be able to construct the assessment matrix and understand what questions to include and exclude the project started with an extensive literature study. The literature study aimed to distinguish important factors in automation as well as understand what difficulties might arise when going from manual production to automated.

The assessment matrix was capable of predicting a suitable way to continue with their design phase and a crude simulation has been done for all automation parts to showcase the results obtained. While the assessment matrix is at an acceptable level of gaining results it was clear during the literature review that there is much potential to go further with a tool like this. This degree project has scraped the surface on all sub-subjects to create a basis to work with during this time limit. but if one is to delve deeper into the subject much more can be found and constructed to create a larger, more accurate assessment matrix.

A sawmill operation uses many techniques, such as debarking, sawing, drying, and planning, to convert unprocessed logs into timber products. To achieve maximum productivity and satisfy client needs, effective coordination, resource management, and quality control are essential.

Sawmill operations may make use of simulation approaches including Discrete Event Simulation (DES), System Dynamics (SD), and Agent-Based Simulation (ABS). By simulating the flow of logs and timber, DES can improve workflows and resource efficiency. SD offers perceptions on feedback loops and causal interactions. ABS provides extensive analysis by modelling individual agents and their interactions.

The investigation began with a thorough data analysis and modelling approach for the sawmill system. There were efforts made to collect qualitative and quantitative data, which were rigorously validated. The system's basic features and processes were then shown in a conceptual model. To test several elements of the system's performance, simulation scenarios were run on the ExtendSim platform using a certified digital model.

The influence of buffer sizes on sawmill throughput and efficiency was studied, and it was discovered that changing buffer sizes enhanced production efficiency and reduced waiting times. Strategic buffer allocation resulted in greater efficacy and lower wait times.

Another set of tests looked at process delays and how they affected sawmill productivity and throughput. Although effective workflow management reduced process durations, the boost in productivity that resulted was not statistically significant.

After looking into input fluctuations (log quantity and quality), the study showed that the best input combinations for increasing sawmill productivity could be determined based on the products that were produced. Controlling input variance turned out to be crucial for raising system efficiency as a whole.

This degree project is aiming to develop a Factory-in-a-Box platform, which refers to a highly mobile, flexible, and agile manufacturing concept to fulfil the current market needs.

The project is conducted by a three-step approach including the definition of the key-requirements, the conception of the platform, and the validation of the platform considering a use-case. The found key-requirements Internal and External Mobility, HSE-Conformity, Reconfigurability and Customization for the Operation Model are being used for the development of the platform concept in the fields of Basic Concep tApproach, Safety, Automation and I4.0.

It can be concluded that the conception of ageneral Factory-in-a-Box platform which stands for itself without considering the use-case is, especially due to the safety requirements regarding the integration of further equipment (inside the EU), not possible. This needs to be addressed via the accompanying business model of the platform and the reconfiguration, safety certification process of the entire use-case related system.

In addition, there are challenges regarding the need for standardized, universal electrical and mechanical interfaces, which are currently in some cases still supplier specific. Universal (e.g. communication) interfaces are one main field of current and future development and research which will enable a leaner integration/reconfiguration process.

Plug and produce manufacturing systems have the potential to significantly increase the efficiency and adaptability of manufacturing pipelines. However, to fully leverage this potential, it is essential to have efficient solutions for hazard safety detection to avoid bottlenecking the efficiency gains. A crucial step in hazard safety detection is effectively relaying information about risks and mitigation strategies to the target operator.

This thesis hence explores the feasibility of using large language models, particularly GPT-4, for generating safety texts in automatic hazard detection systems within smart manufacturing environments. The study focuses on enhancing existing hazard detection outputs into clear, human-readable safety instructions. The overall goal being to optimize this step of the hazard detection procedure.

Through a case study, the research demonstrates that GPT-4 can produce high-quality, fluent, and comprehensive safety texts that address identified hazards effectively. Additionally, a synthetic dataset is created to quantitatively measure how robust GPT-4 is incovering addressed hazards, revealing that it frequently covers all risks. While the generated texts showed high coverage of hazards and were generally well-received by experts, the study also identified areas for improvement, such as the need for better steerability and the inclusion of specific hazard prevention documentation.

Finally, we note that GPT-4 has a high tendency of adding superfluous information that was not requested when generating summaries.

Virtual commissioning is a technique for programming, optimising and verifying industrial automated production, such as robot controllers and programmable logic controllers (PLC), off-line in a simulated environment. Compared with traditional robot off-line programming and simulation, the scope is wider and can include an entire production cell.

Robot simulation is a well-established technique and widely used in industry today, much thanks to the RRS interface that enables simulated robot control systems to be integrated in the simulation software. A more general interface for industrial control system integration is OPC that has been an industrial de facto standard for connection between industrial control systems and regular PCs. State-of-the-art production simulation tools often include the possibility to connect an industrial control system via OPC. However, OPC suffers a major drawback when it comes to production simulation, there is no mechanism that synchronises the industrial control system with the simulation and this could lead to unreliable results from the simulation.

Another obstacle for virtual commissioning is the amount of time that needs to be spent during the simulation model building phase, since virtual commissioning includes more signals. This does not only take more time, but it is also an error prone process that might lead to unreliable results.

In this thesis problems related to the OPC interface and the modelling process are discussed, and suggestions how these issues can be solved are presented so reliable virtual commissioning can be achieved.

Flexible automation cells with rapid product changes are an important competitive advantage for industries today. These cells can increase a company’s productivity and thereby increase their profits. A flexible cell shall be able to handle different products with none or minimal changes to the cell itself. A powerful tool, which can be used to analyse and verify such cells, is discrete event system simulation. Problems such as potential bottlenecks, deadlocks, answers to "what-if" questions and the level of resource utilisation can be gathered. The drawback of discrete event system simulation is that the modelling task is both time consuming and difficult to accomplish. Furthermore, state-of-the-art discrete event system simulation tools that are used in the industry today are not suitable for flexible automation. If the production scenario is changed, e.g. introduction of a new product, the simulation and modelling has to be redone and this is both time consuming and tedious. In this paper a new approach will be presented that enables discrete event simulation models to be generated automatically. The models are generated from information retrieved from a PLM/PDM database system, which is shared among other engineering tools such as robot simulation, CAD and process planning. Hence, when the cell and the database are updated a new model can easily be generated. The database is also connected to the real cell so up-to-date data can be retrieved from the real cell. The model generator described in this paper was implemented and tested in a discrete event system simulation tool and showed promising results. With this approach it is possible to handle flexible automation cells more effectively in a process planning stage.

Simulation based PLC code verification is a part of virtual commissioning, where the control code is verified against a virtual prototype of an application. With today’s general OPC interface it is easy to connect a PLC to a simulation tool for e.g. verification purposes. However, there are some problems with this approach that can lead to an unreliable verification result. In this paper, four major problems with the OPC interface are described, and two possible solutions to the problems are presented: a general IEC 61131-3 based software solution, and a new OPC standard solution

This study aimed to make a specific type of welding, called Friction Stir Welding (FSW), better when welding along curved paths using offline programming. The goal was to reduce two common problems: metal ejection and flashing. Welding along curves might become difficult due to friction, overheating, and down-pressure can cause these issues. The study wanted to find ways to make the welding process smoother and prevent these problems along the curve. To tackle this challenge, the study tried different methods by changing theparameters such as welding speed, tool speed, and force applied. The goal was to find the best combination of settings to reduce metal ejection and flashing.The experiments showed that making some changes to the welding parameter by monitoring the changing path in feedback. By adjusting the force and speed at specific points, the welding process becomes better and good quality. This meant the metal stayed in place and didn't fly out as much. However, the study also found that changing one thing could sometimes cause other problems. For example, adjusting the force could make the robot less stable. So, finding the right balance was important.In the future, using sensors and smart computer programs could help make welding along curves even better. These technologies can keep an eye on the welding process and make quick adjustments to prevent problems. In conclusion, this study worked on making curved welding better by changing welding parameter settings. It found that small changes can make a big difference in reducing metal ejection and flashing. The study also showed that new technologies could make curved welding even smoother and more reliable in the future.

Sensor systems of industrial Autonomous Mobile Robots (ARM) constitute the main part of its information and measurement systems. The purpose of these systems is to form and provide information about the state of objects and processes in the environment and about the robot itself. This information is required for the functioning of the robot.

A sensitive device or a sensor is a primary transducer that reacts to the value to be detected (temperature, pressure, displacement, current, etc.) and converts it into another value, convenient for further use, giving a signal about its presence and intensity. This signal can be of any physical nature, determined by the principle of operation of the sensitive device. It is preferable that it is electrical, since most technical systems in which it will be used are electrical. However, there are systems of a different nature, for example, completely pneumatic, designed to work in conditions that do not allow electricity. In these cases, signals of a different nature must be used.

Requirements for sensor systems substantially depend on the level of the control system at which their information is used. Sensor systems used at control levels operating in real time should have the highest performance, with the inevitable simplification of this information. On the contrary, at the strategic level of controlling the behaviour of the robot, the most complete information is required, possibly to the detriment of performance.

Robotics is a complex and fascinating method for investigation of the surrounding space. A profound moment within the history of most robotics is that the expected value a robot performed a task below the influence of software package or electronics.

This work is devoted to the study of the possibilities of modernization and improvement of the navigation system of mobile robots in order to move in a space with obstacles. The research was carried out at the Industrial In-novation Arena in Skövde (Sweden).

The aim of the work is to develop the most sensitive sensor system of the robot to ensure easy movement in space with movable and fixed obstacles.

In this report, two methods will be used, the main method - Ullmans method and the evaluation method to select the optimal solution - Quality function deployment (QFD) method.

Based on the QFD, it was concluded that the most optimal is to use Simultaneous Localization and Mapping method (SLAM). For a more accurate result, further research in the laboratory will be required.

Nowadays companies aim to become more and more competitive. More production with less wastes of time, energy, matter is the aim of these new tools which are composing the industry 4.0. This new era of digitalization has already started.This paper will present the technologies and concepts necessary to understand the setting up of a digital solution, and its implementation in a factory which would have evolved step by step, by enlarging and creating new sections, with different PLCs.

 It will be explained how the different communication protocols, due to the different brands of PLCs, are creating problems in the communication when it is about implementing a digital solution. It will also propose several potential solutions, and several solutions will be proposed and compared. Also, the reasons of the chosen final solution will be explained. These results have been obtained by consulting several studies concerning the new technologies about industry 4.0, but also by going directly on the field, to notice the problems and applicable solution beyond the theoretical aspect. Interviews on several occasions with the actual leader of the market for the implementation of digital solution is composing a solid and precious testimony about what could be done to implement such solutions for now. Indeed, papers are presenting technologies which could be developed, or already existing, as the one which will revolutionize the interconnectivity, but it seems that this is for a near future, while the actual reality seems to be different. The final decision which has been taken will be explained regarding the methodology followed to solve the problem. It will be presented that a solution which is more appealing is not always the optimal solution, depending of the company configuration, needs and long term goal. This document will present the explaining why and how to determine the corresponding solution for a specific company.