Introduction

Wind power technology is an environmentally sustainable energy source1. The current global wind capacity stands at 743 GW2, a significant rise from 24 GW in 20015. Wind energy accounts for approximately 6% of global energy demand, making it the second-largest renewable energy source after hydropower3. Notably, there has been a remarkable year-over-year increase of 55% in wind power capacity2. Over its operational lifespan, a wind turbine can generate energy equivalent to 23–57 times the energy consumed in its manufacturing4, highlighting its efficiency and long-term environmental benefits.

Wind turbine manufacturers face intense competition to offer turbines with higher rated capacities. The rationale for the design capacity race is the direct correlation between a turbine’s rotor-swept area and the energy it generates. Consequently, wind turbines with higher-rated capacities contribute to the progressive reduction of the levelized cost of energy (LCOE)5. The original equipment manufacturers (OEMs) now offer wind turbines with capacities of up to 17 MW, an eightfold increase from 1998. Notably, the rotor diameter of a 17 MW wind turbine spans 262 m, three times the size of a 2 MW wind turbine6,7 (see Fig. 1).

Fig. 1
Fig. 1
Full size image

Hub height and rated capacity of wind turbines.

The manufacturing of wind turbines primarily relies on manual processes; however, the assembly of large-sized components presents efficiency and safety challenges for human workers. Furthermore, given that wind turbine designs are still in their early stages of maturation8,9,10, frequent design variations make fixed automation solutions infeasible. Intense market competition and high wage rates further fuel the desire for automation within the wind industry11.

A shortage of skilled workers is another challenge that wind turbine manufacturers, like other sectors, face. A study on wind turbine manufacturing value chains11 identified the nacelle and blades as complex components. These high-complexity components of wind turbines are primarily produced in Europe, the United States, and Japan (see Fig. 2). Many countries on this list face three significant workforce-related issues: a declining population, an increasing proportion of senior citizens, and a rising number of elderly individuals in households12. For instance, Japan, Italy, Germany, and Sweden have 20.1%, 19.7%, 18.8%, and 17.2% of their populations aged 65 and above, respectively12.

Collaborative robots (a.k.a. cobots) are a new generation of industrial robots that can operate alongside humans. Cobots can enable flexible automation by combining the best skills of machines and humans13,14. Cobots are safe for humans, mobile, and easy to reconfigure in new situations15. When combined with the required safety technologies16, cobots can form a human-robot team17 scenario to address the challenges of wind turbine manufacturing.

Fig. 2
Fig. 2
Full size image

Diversity in the number and geographic spread of manufacturers by wind turbine components.

The current state of the art in cobot applications comes with its share of challenges. Cobots are predominantly utilized in the manufacturing of small components. These cobots are often designed for specific applications, limiting their versatility and adaptability. Furthermore, their interactions with humans are typically limited or nonexistent. Another challenge lies in the time-consuming changeover and reconfiguration processes in cobot systems. This limitation often leads to purpose-built cobots that are less flexible and agile in responding to evolving manufacturing needs.

This paper proposes the building blocks of a flexible mobile robot assistant (MRA) for the wind industry. A systematic framework to develop the MRA is proposed. The framework involves a digital twin (DT) to test, validate, commission, and reconfigure the physical MRA. An intuitive wearable human-robot interface (HRI) has been developed for fluid interaction on the shop floor. It was observed that the robot assistants reduced person-hour requirements, the HRI method was user-friendly, and the digital twin reduced the development and reconfiguration time for the robot assistants.

The key contributions of this paper are to present a:

  1. 1.

    Generic architecture of a mobile robot assistant for manufacturing applications.

  2. 2.

    A scalable digital-twin-based framework to develop and reconfigure MRAs.

  3. 3.

    Method of human-robot interaction for industrial applications.

  4. 4.

    Prototype of a mobile robot assistant to validate the proposed research in nacelle assembly.

Materials and methods

This chapter presents background knowledge and related work to lay the foundation for the topic.

Components of a wind turbine

Wind turbine technology is complex, encompassing various multidisciplinary technical fields. These include aerodynamics, mechanics, structural dynamics, meteorology, and electrical engineering, all of which play critical roles in addressing different aspects of wind turbines9, from their energy generation to transmission and integration into the power grid.

Wind turbines come in different design configurations, including horizontal-axis and vertical-axis turbines. The design most commonly favored by OEMs is the three-blade, upwind, horizontal-axis wind turbine. In this configuration, the turbine’s rotor and nacelle are positioned atop a sturdy tubular steel tower.

A wind turbine comprises three main components: the tower, nacelle, and blades. A rotor and a hub are within the nacelle, with the blades attached to the hub. A drivetrain-type wind turbine includes components such as a gearbox, shafts, support bearings, a generator, and supplementary machinery18. The rotor harnesses the wind’s kinetic energy, transforming it into rotational energy and transmitting it through the drivetrain to power a generator. The resulting electrical energy can be directed to the load or integrated into the utility grid.

Manufacturing of wind turbines

Current wind turbine manufacturing methods often rely heavily on manual labor. Increasing automation requires large, expensive, and specialized equipment that may not always be economically viable.

Wind turbine blade manufacturing

The manufacturing process for wind turbine blades varies depending on factors such as turbine size, type, materials, and manufacturing techniques. Blades are typically made from composite materials, such as fiberglass reinforced with epoxy resin, chosen for their exceptional strength-to-weight ratio and durability in various environmental conditions. These blades undergo a complex design and development process to ensure strength, durability, and efficiency.

A mold or tooling is developed to create the optimal blade shape for wind utilization and weather endurance. Usually made of metal or composite materials, these molds shape the blade during manufacturing. Layers of composite materials, including fiberglass and resin, are carefully laid up in the mold, following a precise sequence and orientation. Depending on size and complexity, this process can be manual or automated. The blades are later cured in an oven or heated environment to strengthen and maintain their shape.

After curing, the blade’s surface may be painted to protect it from environmental elements. Stringent quality control, such as ultrasound or X-ray inspections, ensures the blade adheres to design specifications and safety standards. After manufacturing and inspection, the blades are transported to the wind turbine assembly site, where they are attached to the rotor hub, completing the wind turbine for operation.

Wind turbine nacelle manufacturing

The wind turbine’s nacelle houses key components, including the generator, gearbox, and control systems. The manufacturing process comprises multiple stages, with variations determined by factors like turbine size and design. Structural elements, such as the frame, base, and housing, are fabricated using welding and machining, thereby forming the nacelle’s core structure. These components are then assembled to construct the primary nacelle frame, which includes structural supports and attachment points for other components.

The generator, gearbox, and electrical systems are assembled inside this nacelle frame. This process involves installing wiring and electrical systems, as well as sensors and control systems, to ensure efficient electricity generation and communication with the central control center. In large wind turbines, hydraulic systems adjust blade pitch and orient the turbine’s direction (yaw control), which are integrated into the nacelle during manufacturing. Safety features, such as fire suppression systems and lightning protection, are also incorporated into the nacelle to protect the turbine from potential hazards.

Related work

Manufacturing automation brings safety, efficiency, quality, and cost-effectiveness19. Industrial automation of physical tasks is primarily attributed to large and bulky industrial robots20 that are immobile, operate in fencing, and are time-consuming to commission and reconfigure21. They are classified as fixed automation solutions offering high production volume, strict separation from humans, and little to no flexibility22.

Humans possess inherent flexibility and adaptability, enabling them to adjust to changing scenarios23. Leveraging this innate flexibility and combining it with automation systems has given rise to human-robot collaboration within manufacturing settings24. Cobots, have opened up new avenues for automation by tackling previously challenging tasks using conventional robotics15,25.

The literature showcases a range of cobot applications, including pick-and-place operations, assembly tasks, welding, inspection processes, and packing26. Moreover, cobots have been investigated as a solution for quickly repurposing factories during emergencies13. However, it’s worth noting that most cobot applications have been observed in small component manufacturing, such as electronics, appliances, and electronic actuators15,16.

Mobile robots are another form of collaborative robots27. They have been used for material handling applications in manufacturing settings28. Mobile robots are well-suited for tasks that require frequent human interaction and mobility. They find applications in factories, military operations, healthcare, search and rescue, security, and homes15.

A combination of an articulated robotic arm and a mobile robot is often called a robot assistant29. Kirova30 reviewed MRAs in automotive manufacturing and assembly. The study also presented the requirements of an MRA in automotive manufacturing (Table 1). Yamazaki12 discussed a home MRA for the aging population. An example of a mobile robot assistant, “Little Helper,” operating in a laboratory environment, was presented by31.

Table 1 General criteria for industrial use of assistive robotics in the automotive industry30.
Full size table

Aircraft manufacturing has similarities with wind turbine manufacturing. The Air-Cobot project32 reported the development of a collaborative MRA for airport-based routine aircraft inspection before takeoff. Robotic assistance in aircraft assembly has been reported by33, however, the work concerns humanoid robotic assistants with bipedal walking. The available literature has limited application of human-robot collaboration (HRC) in large component manufacturing34. A key challenge is the allowable payload capacity in collaborative applications.

TwinHRC: digital twin for human-robot collaboration

An MRA is typically a reprogrammable device equipped with mobility to navigate different locations and a robotic arm to perform a predefined set of tasks. To function as a “coworker,” an MRA must replicate a range of human skills for effective teamwork. This includes receiving and processing information and carrying out tasks to collaborate with human colleagues.

The proposed TwinHRC framework enhances system flexibility by leveraging digital twin technology for design, verification, and validation, and by modularizing software and hardware, ensuring scalability across applications at multiple levels. The framework follows a structured progression through six distinct phases, guiding the development of the robotic assistant (see Fig. 3).

  1. a.

    Problem identification and objectives: This process entails identifying the requisite elements for developing an MRA. It involves documenting the requirements and establishing success criteria for an efficient MRA. Additionally, this phase includes estimating the expected business value, economic impact, and return on investment.

  2. b.

    System design: This stage encompasses the system-level design of the robot assistant, which includes defining robot manipulators and related hardware components, either custom designs or selection from available off-the-shelf options. The workspace design is aligned with the robot’s capabilities and is verified and validated. Moreover, a logical workflow is formed, driven by the automation potential of tasks, takt time, and resource availability.

  3. c.

    Digital twin for design and control: The DT layer supports, validates, and monitors the robot system’s design, control, and performance. It runs in parallel with the design, development, commissioning, and operations phases. This layer simulates the robotic system’s design, assesses its performance, and conducts analyses that replicate the operational aspects of an assembly line. The control program, function blocks, and automation logic for controlling the robot system are also embedded into the DT.

Fig. 3
Fig. 3
Full size image

TwinHRC framework to develop a robotic system in manufacturing.

  1. d.

    Development: This stage involves constructing the robot system according to the design specifications established during the design phase. Safety risk assessments are conducted, and safety systems are evaluated to ensure a secure working environment. Given the emphasis on regular human-robot interaction, the system integrates an interaction method to facilitate fluid HRC.

  2. e.

    Commissioning: The developed robot system is commissioned for serial production. The DT continues to play a vital role in supporting this stage. Since the system has undergone thorough evaluation in the virtual environment, the physical system is commissioned to mirror the design and processes established in the DT.

  3. f.

    Operations: The developed system is continuously monitored during serial production for performance optimization. User studies are conducted, and PDCA (Plan-Do-Check-Act) cycles are implemented to refine and constantly optimize the system’s performance. Overall Equipment Effectiveness (OEE) is another measure that can be integrated into the analysis engine to monitor and maximize robot system performance. This iterative approach helps fine-tune the robot assistant’s capabilities to align with production requirements and user feedback.

Architecture of the proposed MRA

This section presents a generic architecture of a flexible MRA tailored for manufacturing applications (Fig. 4). The key characteristics of the MRA include autonomous mobility, manipulation capabilities, precise localization, adaptability across tasks, intuitive human-robot interaction, and safety. A description of these characteristics is presented below:

Fig. 4
Fig. 4
Full size image

Architecture of the proposed mobile robot assistant.

  1. a.

    Robot system: A robotic manipulator is to execute the assigned assembly tasks. Essential factors in choosing the appropriate manipulator include its reach, payload capacity, and repeatability. Dual-arm robots can also offer greater flexibility, enabling more intricate operations. When mounted on an autonomous mobile robot, the manipulator enables the entire robotic system to navigate autonomously. Additionally, additive manufacturing can make the robot tool more flexible and cost-efficient. This adaptability empowers it to handle various tasks and accommodate design alterations.

  2. b.

    Reconfigurability: Manufacturing environments often exhibit a semi-structured nature, encompassing tasks that range from simple to complex. A significant number of these tasks offer considerable potential for automation. An MRA should not be bound to a specific task; instead, it should be designed for reconfigurability, enabling it to adapt to various functions seamlessly without requiring significant manual intervention.

To facilitate this reconfigurability, the robot must be capable of changing its end-of-arm tool, creating or loading new programs, and defining new robot trajectories. Integrating an automated tool-changing mechanism can enable the robot to switch tools autonomously, allowing it to perform a wider range of tasks. Additionally, a modular plug-and-play architecture streamlines adding or removing hardware components, enhancing the robot’s flexibility and adaptability to evolving manufacturing needs.

  1. c.

    Mobility: Mobility is another needed characteristic that allows the robot to navigate to its designated location without requiring manual guidance. This requires the robot to maneuver autonomously, precisely, and safely to the desired position. Autonomous mobile robots have seen remarkable progress in recent years, and their potential can be expanded by integrating flexible robotic arms to perform a variety of tasks.

  2. d.

    Localization: The localization of the robot assistant is necessary to enable the robot to understand its surroundings, thereby allowing it to “know its world”. The appropriate localization technique should encompass accuracy, repeatability, cost-effectiveness, and reconfigurability. Moreover, it should dynamically adapt to the environment as the robot navigates to various locations, enabling it to determine its reference (home) position.

  3. e.

    Human interface: Operators may need to interact with an MRA during task execution, allowing them to issue commands for task progression, restart processes, or halt the robot as needed. A wireless interface is required to allow the operator to move within the workspace. Additionally, the interface should prioritize simplicity and user-friendliness.

  4. f.

    Task execution and synchronization: To coordinate the key functions of an MRA, such as mobility, manipulation, reconfigurability, and human interaction, Behavior Trees (BTs) can provide a structured and modular framework for task execution. They decompose high-level goals (e.g., “assemble a part”) into modular sub-tasks (e.g., navigate, localize, grasp, and place), where each sub-task is represented as a node35. Execution passes through the tree based on logic conditions. The tasks can be executed in sequence or in parallel, depending on the logic. This modularity simplifies programming and allows reconfiguration to add new tasks without significant alterations to existing logic.

In addition to task execution, BTs enable the synchronization of concurrent actions and the integration of human intervention36. For example, while the MRA executes a pick-and-place action, the mobility system can monitor navigation status, and the human interface can remain responsive to operator commands. Synchronization nodes (e.g., sequence, selector, parallel) make the tasks interdependent. At the same time, human-in-the-loop control allows operators to interrupt, pause, or reprioritize tasks without disrupting the overall workflow37.

The development process is described in the following chapter using a use case.

An industrial case

A case study from a leading wind turbine manufacturer illustrates the design and development of an adaptable MRA. This case study specifically focuses on the serial production of more than 10 MW offshore wind turbines.

Problem description

The bed frame is a critical structural component within the wind turbine’s nacelle, supporting various subassemblies. Its design also allows the rotor to rotate, helping compensate for changes in wind direction during operation. Electric cables are deployed inside and outside the bed frame to facilitate electrical connectivity. In the given example, seventeen cable trays are installed internally, and thirty-six are installed on the exterior of the bed frame. These cable trays are attached using magnetic clamps.

The cable trays were manually assembled in the example presented. This manual assembly process required a human operator to stand on an elevated work platform (EWP) for outer assembly and work within the confines of the bed frame for internal assembly. While technical drawings were available to guide the placement of each cable tray, manually measuring linear distances was challenging due to the bed frame’s curved surface, as shown in Fig. 5.

Fig. 5
Fig. 5
Full size image

Dimensions of the cable trays.

Manual assembly posed the following challenges:

  • Time-consuming measurements: Taking manual measurements to identify assembly location on bed frames was a repetitive process, consuming significant time.

  • Depth perception difficulty: Operators working near the bedframe struggled with depth perception, making it challenging to stay aligned with reference points during assembly.

  • Complex cable tray interconnections: Wired interconnections between cable trays complicate their assembly handling.

  • Cumulative dimensional errors: Minor misplacements could accumulate, leading to substantial errors in the final assembly.

Manual assembly often resulted in notable variations in component placement within the final assemblies, sometimes exceeding ± 100 mm. When electric cables are installed in these cable trays, they can become too short. This situation necessitated repositioning and realigning cable trays, which could entail several hours of rework.

Design of the robot assistant

A robotic assistant was proposed to enhance worker efficiency by projecting target assembly locations onto the bedframe surface using laser beams. A laser device mounted at the robot’s tool post projected laser markings at each cable tray location, eliminating the need for manual measurements. The robot was supposed to systematically target each assembly location until all cable trays were assembled. Operator interaction with the robot was facilitated through a wearable device. This robotic system was designed to deliver precision and consistency while relieving operators of repetitive tasks, as shown in Fig. 6. The MRA system comprises three main components: a robot manipulator, an autonomous mobile robot, and a flexible EOAT at the robot manipulator’s tool post (Table 2).

Fig. 6
Fig. 6
Full size image

Using an MRA in wind turbine manufacturing.

Table 2 Capabilities of the mobile robot assistant.
Full size table

Mobile robot assistant (MRA) system

The robot manipulator was Doosan M0609, a collaborative robot with a force-limiting body. This robot possessed six degrees of freedom (DoF), offering a reach of 900 mm and a payload capacity of up to 6 kg. Additionally, the robot had a programmable cockpit on its arm, providing a convenient and rapid interface for operator interaction. The robot manipulator was mounted on top of an autonomous mobile robot. The mobile robot enabled autonomous maneuvering of the robot system and allowed for the storage of tools and peripherals.

The robot tool was designed to be flexible and cost-effective (Fig. 7). This flexibility in tool interchanging enabled it to perform multiple tasks and adapt to future product design changes. This modular tool was manufactured using additive manufacturing. The tool could hold four lasers and accommodate a calibration camera when needed. A snap mechanism (for quick attachment and detachment) was integrated to simplify fastening the laser tool to the robot. As a result, the tool can be attached or detached using a snap-fit mechanism, eliminating the need for additional fasteners.

Fig. 7
Fig. 7
Full size image

Laser tool for the robot assistant. The right image shows the tool design modeled in SolidWorks 2021 (https://www.solidworks.com/)38. The left image shows the tool developed using additive manufacturing.

Process planning

The robot’s ability to move around the workspace was essential for accurately guiding the worker to assembly locations, a capability enabled by the platform’s mobility. To ensure precision, the robot strategically navigated to multiple locations within the workspace, accounting for both the laser range and its reach, enabling it to perform its tasks effectively. For prototyping, a manually movable platform was used in place of a mobile robot. The integration of a mobile robot will be investigated in future work.

Mobility and localization

Localization is critical for calibrating the robot to its environment, thus ensuring precise task execution. A robust localization system is employed to configure the MRA for each new sub-assembly on the production line. Key considerations in this context include the time required for calibration, the level of accuracy achieved, and the associated costs of the calibration system. The robot system is designed to undergo localization for internal and outer cable trays, ensuring accuracy across all assembly areas.

Initially, a mechanical fixture-based localization method was adopted. The mechanical fixtures are aligned with the fixturing features on the bed frame, which remain constant to ensure repeatability. However, an intelligent and automated system requires vision-based localization. Various techniques for vision-based localization exist, including marker-based, 3D camera-based, and laser-based approaches.

Digital twin to develop, control, and monitor

Typically, the design of a product like a wind turbine is finalized well before planning the manufacturing systems. In this context, the MRA was required to rely on the bed frame’s design and CAD models as references. The CAD data was stored in STEP (Standard for the Exchange of Product Data) format and imported into the RoboDK simulation software. The cable tray positions were derived from these CAD models and subsequently entered into the robot’s program. This process created a high-fidelity simulation during the design phase, enabling accurate visualization and analysis of the robot’s operations (Fig. 8).

Fig. 8
Fig. 8
Full size image

Simulation of the robot assistant for task verification developed in tecnomatix process simulate.

A significant challenge was modeling the laser beam. The robotic simulation software could not model light waves (i.e., wavelength, beam profile, and propagation medium, which define a laser beam’s properties). Therefore, a solid model identical to a laser beam was used to simplify the process, but it offered limited accuracy. This problem can be solved using gaming engines that model light waves and robotic processes.

An offset was also observed between the simulated and actual laser beams. To address this challenge, manual measurements were obtained directly from the physical laser and were integrated into the simulation. This process ensured that the simulation accurately mirrored the laser’s real-world behavior and offset, enabling a more precise representation of the laser’s actions in the virtual environment.

The simulated robot was connected to the physical robot in real time via an Ethernet connection. In the simulation, the IP address of the physical robot was specified to establish a TCP/IP connection with the controller, enabling communication between the virtual model and the actual hardware (Fig. 9). Once the connection was established, the robot was disengaged from its teach pendant, and control was handed over to the simulation. Within this integrated environment, it became possible to transfer joint positions from the simulation to the real robot.

For instance, robot positions from the simulation could be sent to the physical robot and vice versa. This bidirectional data exchange facilitates synchronization between virtual and real-world robots, streamlining programming, testing, and fine-tuning of robot actions in a controlled digital environment before implementation on the physical robot.

Fig. 9
Fig. 9
Full size image

A simulation-based digital twin of the robot assistant for online programming.

Additionally, the simulation robot was integrated with the Doosan robot virtual controller, allowing for the execution of actual joint specifications and programming languages. This was achieved by developing a dedicated post-processor within the simulation tailored for Doosan Robotics. A robot post-processor in the simulation defines how robot programs should be generated to be compatible with the unique specifications of a particular robot controller, ensuring that the robot’s actions are accurately represented and executed.

Intuitive human-robot interaction

An intuitive human-machine interface (HMI) ensures effective human-robot interaction on the factory floor. This HMI should enable operators to easily initiate, halt, pause, or reset tasks. The HMI must be designed with considerations for intuitiveness, cost-effectiveness, and compliance with safety standards.

Fig. 10
Fig. 10
Full size image

Developed prototypes for HRI interaction.

Various HMI prototypes were developed, including smartwatches, remote controls, and smartphones, all designed to provide a wireless, straightforward interface (as depicted in Fig. 10). These prototypes were tested on-site with operators, and based on their feedback, the smartwatch interface proved to be the most viable option. The smartwatch interface was integrated with an intelligent switch (Sonoff 4CH Pro), which features four channels (I/O ports) that can be controlled via the four buttons on the smartwatch (Fig. 11).

Fig. 11
Fig. 11
Full size image

Process architecture of smartwatch-based human-robot interaction.

Automation of assembly tasks

The system required a laser device capable of projecting a laser beam effectively onto the bed frames. The laser markings needed to remain visible from 8 to 10 m, guiding operators to specific assembly locations. To fulfill this requirement, the Bosch Quigo laser was chosen. This device, typically used as a cost-effective laser leveling tool for household maintenance, proved suitable. Two of these laser devices were attached to the robot tool.

It’s important to note that every laser device has a degree of error in its laser beam projection. This means the laser projection is not perfectly parallel to the device’s axis; instead, it forms an angle. Since the robot’s programming relies on a simulated digital twin, a misalignment may occur between the actual and simulated laser projections. A compensation error procedure was performed to rectify the misalignment and ensure the laser beam was projected accurately to the intended locations.

Reconfigurability of the robot assistant

The robot system’s versatility is a key objective, as it is intended to perform multiple tasks. Reconfiguring the robot involves replacing tools and adjusting its program. Another application scenario for the robot was identified: assembling bolts. The robot is transferred to a robotic cell when it is not engaged in cable tray assembly. An automated tool-changing mechanism has been developed to enable the robot to switch tools and perform bolt-assembly tasks. This flexibility is a significant advantage of the robot assistant. Given that wind turbine manufacturing predominantly relies on manual labor, there is substantial potential for robots to automate repetitive tasks across various aspects of the production process (Fig. 12).

Fig. 12
Fig. 12
Full size image

Experimental results of using a robot assistant in assembly.

Test results

The tests showed overall compliance with the desired objectives. The test results are evaluated based on accuracy, repeatability, time reduction, and ease of operation.

Improved process quality and repeatability

The projection of cable tray locations and orientations onto the bedframe was within the desired accuracy range of 5 mm. No instances of cable length shortening were observed at the end of the assembly line in 20 experiments (Fig. 13). The laser beam was visible on the bed frame, and the worker could identify its locations and angles.

Fig. 13
Fig. 13
Full size image

Results from the test. The robot is beaming the first horizontal cable tray position (left) and the first vertical cable tray position (right).

Reduced person-hours

Manual assembly of cable trays typically takes 40 min and often requires frequent manual adjustments to ensure proper alignment, relying on visual inspection. With MRA, the processing time was reduced to about 25 min—a 37% decrease—without requiring adjustments. Implementing an MRA improved consistency in cable length during bed frame production.

Social acceptability

Various studies in the literature have discussed the social acceptability of human-robot collaboration. It has been shown that concerns about job displacement can be mitigated when system development processes are designed to be transparent and inclusive. Observations also indicated that the assembly process became more efficient, eliminating the need for manual measurements or visual adjustments. Additionally, the robot’s control via a smartwatch interface demonstrated an intuitive and user-friendly design. These findings emphasize the robot’s potential to enhance assembly processes.

Discussion

The limitations of the project and further work are presented below:

  1. a.

    Robot calibration: Calibrating and localizing the robot were pivotal aspects of the project. Currently, a manual localization technique is employed using mechanical fixtures. However, repeated use of these fixtures is not recommended due to their size and weight, as well as the potential risks they pose to the robot system, such as damaging the robot platform or causing accidental collisions.

As a solution, the project aims to implement a vision-based calibration system to enhance convenience and durability. This system would leverage visual data and image processing techniques to precisely determine the robot’s position and orientation, eliminating the need for cumbersome mechanical fixtures and mitigating potential risks associated with their use.

  1. b.

    Alignment with safety standards: Care has been exercised in the system’s design to ensure compliance with safety regulations outlined in the ISO/TS 15066:2016 safety standard. This includes implementing safety features such as a safety-monitored stop, force limiting on the robot’s body, and avoiding sharp edges in the robot’s tooling. However, any alteration to the physical workspace will require a comprehensive safety risk assessment. This assessment will evaluate the robot’s safety implications for the system and its intended application.

  2. c.

    Mobility: The MRA design was initially conceived as an autonomous system. However, for practical reasons, the developed prototype does not currently incorporate autonomous mobility features. Fortunately, the market offers a variety of mobile robots that can be utilized for this purpose. Future testing and validation efforts will be essential to confirm the viability and effectiveness of autonomous mobility scenarios within the robot assistant’s functionality. This approach will help ensure the full potential of autonomous mobility is realized in practical applications.

  3. d.

    Reducing the overall cost of wind energy: Robotics, particularly cobots, hold significant potential in large-component manufacturing to minimize person-hours and enhance productivity. There is a growing need for comprehensive studies to explore additional applications of cobots in wind turbine manufacturing and assembly processes. Moreover, it’s essential to investigate whether cobot technology can reduce energy production costs in the wind industry. These studies can provide valuable insights into how cobots can be further integrated into the wind turbine manufacturing process.

Conclusion

The article emphasizes the importance of flexible automation in manufacturing large-sized components, particularly within the wind turbine industry. It proposes a scalable architecture for an MRA designed for wind turbine manufacturing. The critical attributes identified for an MRA are adaptability, user-friendly human interface, and mobility. Notably, the introduction of laser projection guidance for operators has not only simplified the assembly process but also reduced assembly times. Operators expressed satisfaction with the robotic assistant and reported feeling comfortable working alongside it.

The article identifies several fundamental challenges in the integration of mobile robot assistance, including the cost of safety technologies, the need for reconfigurability, the localization of mobile robots, and the importance of intuitive, user-friendly HMIs. Furthermore, the article emphasizes that because wind turbine manufacturing has a progressive assembly line layout, idle time can occur when the robot is not in use. Therefore, the designed cobot is flexible and can be reconfigured for alternate tasks by changing the end-of-arm tooling and switching to a new robot program.

Although cobots are constrained by safety requirements, which limit the range of tasks they can perform, the article contends that they should not be dismissed from large-component manufacturing. It argued that despite their low payload capacity, cobots can still help automate tasks and reduce wind turbine manufacturing costs. Lastly, the article highlights the role that clean and renewable energy, enabled by technologies such as wind turbines, will play in achieving the United Nations’ Sustainable Development Goals (SDGs). Access to sustainable energy sources is deemed integral to realizing these global objectives.