Robotic glass bonding process in built-in Cooker Hood production: system design and efficiency analysis

2.1. End-effector and dispensing system

In robotic adhesive processes, the end-effector serves not only as the physical contact point of the system but also as the control center for the hydrodynamic parameters that determine process quality. Liquid dispensing in automation systems fundamentally relies on the complex relationship between fluid rheology, nozzle geometry, and applied pressure [15]. According to the theoretical framework established by [16] and [17], the volumetric flow rate ($Q$) in automated dispensing systems is directly related to the applied pneumatic pressure ($P$), the viscosity of the fluid ($\eta$), and the internal diameter of the nozzle ($d$). Although an increase in pressure linearly increases the flow rate under ideal conditions according to the Hagen-Poiseuille law, in thixotropic (fluids whose viscosity decreases under shear stress) and high-viscosity materials such as silicone, this relationship becomes non-linear, evolving into a process variable that is difficult to manage [18].

Such high-precision systems offer significant advantages compared to manual application, including micron-level repeatability and material consumption savings [15]. However, the sustainability of these advantages depends on overcoming fundamental challenges within the industrial environment. A primary challenge frequently highlighted in the literature is the instantaneous alteration of adhesive viscosity induced by ambient temperature fluctuations [8]. It is noted that even a slight decrease in temperature increases viscosity, thereby raising the flow resistance at the nozzle tip, which leads to thinning or interruptions in the bead profile [6], [8], [17]. Another critical challenge is the issue of dragging and accumulation [6]. If the dispensing rate remains constant while the robot decelerates during corner turns, silicone accumulation occurs at the corners; conversely, during sudden accelerations, the bead thins or breaks [7].

To manage these hydrodynamic and kinematic challenges, the control architecture of the system is structured around the concept of the Tool Center Point (TCP) [12], [13], [19]. The TCP is a virtual point at the very tip of the nozzle that defines the robot’s position, orientation, and velocity in space. In robotic trajectory planning, regardless of the joint velocities of the robot arm, the primary critical parameter is the tangential velocity of the TCP on the glass surface. For high-quality sealing, seamless synchronization is required between the TCP velocity ($V_{tcp}$) and the volumetric flow rate ($Q$) of the silicone exiting the dispenser [7], [13]. The system utilizes a velocity-adaptive control loop that proportionally reduces the flow rate at corner points where $V_{tcp}$ decreases and increases it to maximum speed on straight lines. The structure of the hybrid end-effector designed in light of these theoretical and practical requirements is detailed in Fig. 1. The most prominent structural feature in the visual is the manner in which the dispensing nozzle and the handling system are integrated.

In robotic dispensing systems for applications such as bonding, having a stable mechanical structure is essential for ensuring flow precision as well as the safe handling of heavy and fragile loads under dynamic conditions [13]. The pneumatic nozzle seen in Fig. 1 is equipped with a high-precision valve system capable of regulating pressure against viscosity changes. Additionally, another innovative aspect of the design is the group of vacuum pads placed symmetrically around the nozzle. When the robot reaches high accelerations during PTP (Point-to-Point/Joint) movements to reduce production time, significant inertia forces are generated on the glass panel [11-13]. The multiple vacuum pad arrangement detailed in Fig. 1 prevents the glass from slipping or breaking under stress by spreading these forces over a large surface area.

Fig. 1Integrated 6-axis robot manipulator and end-effector structure. Photo by Altuğ Pirselimoğlu, Amasya, Türkiye, June 12, 2025

a) 6-axis robot

b) Positioning of vacuum pads

2.2. Motion control and programming architecture

The established system consists of a 6-axis FANUC M-710iC/70 industrial robot and a high-viscosity adhesive dispensing unit. The robot has a maximum linear speed of 2500 mm/s and a repeatability accuracy of ±0.03 mm [12]. In the system, a barrel pump with a pneumatic drive and a precision regulator that prevents pressure fluctuations are used for the continuous feeding of high-viscosity silicone (Shore A hardness: 25-30). To ensure that the 8.20 ml dosage remains constant in each cycle (45 s), the system pressure is managed via a proportional valve integrated with the robot controller. In this study, a kinematic approach defined in the literature as “Stationary Tool-Moving Part” was adopted. In this method, vacuum pads carrying the glass panel are integrated into the end-effector of the robot, while the silicone nozzle (Tool) is fixed at a stationary position in space. This configuration is based on the Remote TCP principle stated by [20], which specifically prevents hose twisting during the processing of large parts and increases trajectory precision.

Trajectory planning and motion control of the system are provided by a hybrid architecture running on the FANUC R-30iB Plus control unit. This architecture relies on the integration of signal processing algorithms managed by the Teach Pendant command set, which provides operational flexibility. Due to the nature of the bonding process, two different motion strategies (interpolation types) were devised with a focus on speed and precision.

PTP interpolation was used for air-cut movements, such as the robot’s approach to the operation area from the Home position, part transfer, and retraction to the safe zone after processing, where trajectory precision is not critical. In PTP motion, the robot controller calculates the shortest path between the start and end points not as a linear line in Cartesian space, but as a curve that the robot joint angles (J1-J6) will complete with the minimum energy and in the shortest time. In this type of motion, the robot's tip (TCP) traces an unplanned arc and the motors work at maximum speed and acceleration. This strategy was the primary factor in reducing total cycle times by 37.8 % by minimizing non-value-added transfer times (Fig. 2(a)).

In the trajectory planning of industrial robots, it is stated that linear interpolation performed in Cartesian space ensures that the robot tip (TCP) moves with a constant velocity profile [20], [21]. For this purpose, LIN (Linear) movement commands were used during the active process where silicone is applied to the glass surface to guarantee bonding quality. Furthermore, as emphasized by [9], [22], synchronization between the flow rate and the robot speed is possible thanks to the constant speed provided by LIN movement, and a uniform bead width can be achieved. In LIN movement, the control unit continuously updates complex inverse kinematics calculations necessary for the TCP to advance along a perfect line between the start and end points (Fig. 2(b)). In this way, a constant speed is achieved by preventing the robot from accelerating or decelerating during turns or axis movements.

Fig. 2Joint and linear interpolation: a) high-speed transfer motion and trajectory curvature performed via PTP; b) constant-velocity silicone application trajectory achieved via LIN

a) TCP Trajectory (curved). Fast but uncontrolled trajectory

b) LIN Trajectory (linear). Precise and controlled trajectory ensuring uniform silicone thickness

Start-stop motions of the robot at the corners of rectangular cooker hood glasses may lead to silicone accumulation. To address this issue, the FANUC CNT (Continuous) termination type was employed within the system. Instead of reaching the exact corner point, the robot rounds the corner within a predefined tolerance radius. This approach maintains the integrity of the sealing bead by ensuring fluid motion at the corners and preventing the robot's velocity from dropping to zero.

3.1. Workstation design and coexistence study

In industrial assembly systems, semi-automated assembly cells, which provide a balance between full automation and manual production, emerge in the literature as a hybrid model that optimizes both flexibility and precision. Within this operating principle, defined as “Coexistence” by [23], humans and robots share the same production goal; however, in compliance with occupational health and safety standards, they perform their actions in different time intervals or segregated spatial zones. In the proposed system, the cognitive flexibility of the operator (quality control, part feeding) is integrated with the kinematic precision of the industrial robot (trajectory tracking, repeatability). The physical layout of the cell is configured based on five primary operational zones, as illustrated in Fig. 3, to minimize material flow while maximizing ergonomic requirements.

The process flow and the interactive synchronous operation between these zones proceed as follows:

Raw Material Feeding and Quality Control (Zones 1-2): The process initiates with the operator loading the raw glass panel onto station (1) and the metal body components (sheet metal, brackets, etc.) onto station (2) within the safety zone. This phase represents the only “human-centric” step of the system; the operator visually inspects the glass surface for potential scratches or screen-printing defects, thereby preventing defective parts from entering the line at the source.

Robotic Manipulation and Preparation: Upon the operator issuing the “Cycle Start” command, the safety barriers are activated, and the FANUC M-710iC/70 robot is engaged. The robot retrieves the glass from area (1) using a vacuum-based hybrid end-effector.

Precision Dispensing (Zone 3): Following the initial handling, the glass panel is transferred to the stationary silicone nozzle located in zone (3) by the manipulator. Here, silicone is applied with micron-level precision to the surfaces where the glass and metal will be joined, utilizing the LIN interpolation technique detailed in Section 2.3. The stationary nozzle and mobile robot architecture (Remote TCP) allow the robot to track complex geometries with greater agility.

Assembly and Curing (Zone 4): Once the silicone application is complete, the part is advanced to the main assembly fixture at station (4) for final integration. The glass panel is aligned and placed onto the metal sheets previously positioned in area (2) by the operator. The pneumatic pistons of the fixture are then activated to apply the pressure required for the adhesive to achieve its initial bond strength.

Discharge and End of Cycle (Zone 5): Once the pressing duration is completed, the fixture opens. The robotic arm or an automated transfer unit moves the assembled product to the finished product output area at station (5). At this point, the system deactivates the light curtain signals to allow the operator to safely retrieve the part and prepares for the subsequent cycle.

This spatial design ensures full compliance with the safety standards for guarded robotic systems detailed by, [13] and [10] while simultaneously supporting lean manufacturing principles by minimizing operator walking distances.

Fig. 3Layout and safety zones of the semi-automated assembly cell

The cell design consists of the following primary components:

– Robotic Manipulator: A 6-axis FANUC M-710iC/70 robot.

– Assembly Fixture: A pneumatically actuated clamping unit ensuring high-precision positioning of glass and sheet metal parts.

– Safety Architecture: Operator safety is ensured through physical barriers and light curtains.

3.2. Operational workflow

With appropriate planning, it is observed that quality is maximized in manual-loading automated assembly lines, which possess lower investment costs compared to fully autonomous systems [24]. In this regard, the robot station features a structure that integrates the flexibility of manual preparation with the stability of high-accuracy, robot-assisted assembly. The process steps depicting the robot-human interaction at the workstation are presented in Fig. 4. At the commencement of the process, the raw glass panel intended for assembly is visually inspected by the operator; if scratches, screen-printing defects, or cracks are identified, the part is rejected as scrap prior to entering the line. Defect-free glass panels are placed in the robot reference area (Fig. 4(a)), and the metal connecting plates and plastic components for the assembly are positioned on the aluminum assembly fixture according to the product recipe, as illustrated in Fig. 4(b). This step, critical for ensuring geometric accuracy, is performed by the operator.

Fig. 4Coexistence process steps. Photo by Altuğ Pirselimoğlu, Amasya, Türkiye, June 12, 2025

a) Placement of the product glass into the feeding area

b) Glass fixture with metal plate and plastics

c) Silicone application via remote TCP

d) Assembly and Pressing.

In the subsequent phase, after the operator moves to the safety zone and issues the start command, the robot grips the glass panel via vacuum using the hybrid end-effector detailed in Fig. 1. The gripped glass is moved along the edge trajectory beneath the stationary nozzle shown in Fig. 4(c) for silicone application. As observed, rather than the conventional method where the silicone gun is mounted on the robot arm, a remote TCP method is employed where the nozzle remains stationary while the part is mobile. This approach prevents the entanglement of robot cables or hoses and ensures a more fluid trajectory tracking. In the following stage, the glass panel, to which silicone has been precisely applied, is transferred to the main assembly fixture shown in Fig. 4(d) for the completion of the assembly process. At this stage, the robot aligns and places the glass panel onto the metal sheets previously positioned by the operator with a precision of ±0.05 mm to prevent the glass from slipping upon contact with the sheets, a vertical ($Z$-axis) approach profile is followed using LIN motion commands within the robot program. Finally, during the assembly phase, pneumatic pistons on the fixture are activated to apply controlled pressure for 3 seconds, ensuring uniform distribution of the adhesive and the development of initial bond strength. Upon completion of the process, the robot retracts to the safe home position, and a visual warning is provided to the operator via a signal lamp to indicate that the product can be removed from the finished area.

The operational cycle of the robotic cell is executed within the framework of a logical flow algorithm schematized in Fig. 5, running on the R-30iB Plus control unit. The algorithm follows a closed-loop structure that ensures operator safety and process consistency.

Fig. 5Control flow diagram of the robotic glass bonding process

The process in the flow diagram presented in Fig. 5 is triggered by the “Start” step and continues with “Manual Loading”, where the operator places the raw material onto the fixture. To proceed to the autonomous operation phase, approval must be obtained from the “Loading Verification” decision block. At this stage, where sensors on the fixture verify the metal and plastic components of the product, if a loading non-conformity (N) is detected, the system is directed to the “Adjust Loading” step, and corrective action is requested from the operator. Following the adjustment, the process is re-verified to account for potential disruptions (Operator Approval); if the operator approves (Y), the process resumes. In the event of an unforeseen issue (fixture or sensor-originated failures) where operator approval is not granted (N), the system executes the “Generate Error Code” block, creates a fault report, and enters the “STOP” state to await service intervention.

Upon successful verification of the loading (Y), the robot reaches the operation point with a “Robot Positioning” motion, and the “Gun ON” command is issued. Before application, the critical “Silicone Flow Control” step is performed. If a flow problem is detected (N), the system is directed to the “Generate Error Code” block and halted. In the normal scenario where flow is verified (Y), the robot performs the “Silicone Application” along the defined trajectory and switches to the “Gun OFF” position upon completion. During the assembly phase, parts are joined via the “Automated Assembly” step, and the fixture pistons are locked by setting “Clamps ON”. To ensure bonding strength, controlled pressure is applied during the “Pressurized Waiting” period (3 s). At the end of the duration, the pistons are released with the “Clamps OFF” command, and the operator is notified via the “Process Completion Signal”. After the robot retracts to the safe area with a “Return to Home” motion, the operator performs the “Unload Finished Product” operation. At the end of the cycle, the system awaits approval from the operator for the next work cycle via the “Confirmation for Next Cycle” decision block. If the operator grants approval (Y), the system loops back to the “Start” block for a new cycle; if not (N), the system proceeds to the “Finish” state and terminates operation.

3.3. Flow control and mathematical modeling

To achieve a high-quality sealing process, the cross-sectional area of the silicone bead (Bead Geometry) must remain constant regardless of variations in robot speed. An Open-Loop control architecture was implemented in the system to facilitate real-time synchronization between the robot's tangential velocity ($V_{TCP}$) and the volumetric flow rate ($Q$). The fluid dynamics of the dispensing process were structured based on the model proposed by [15]. Accordingly, the relationship between the nozzle's volumetric flow rate (Nozzle), applied pressure ($P$), and viscosity ($\eta$) is expressed in Eq. (1):

where, $d$ represents the nozzle inner diameter (mm), $\mathrm{\Delta }P$ denotes the pressure difference, $\eta$ indicates the fluid viscosity, and $L$ corresponds to the nozzle length. However, as viscosity and nozzle geometry are assumed to be constant in industrial applications, the control algorithm is reduced to a proportional relationship between robot speed and flow rate. According to the constant bead width principle detailed by [22], the flow rate must be increased linearly as the robot speed increases. The control structure implemented in the system is integrated into the robot controller via the transfer function provided in Eq. (2):

where, $V_{analog}$ represents the 0-10 V control signal transmitted to the dosing unit, $v_{robot}$ denotes the instantaneous tangential velocity of the robot on the glass surface (mm/s), $K_p$ signifies the calibration coefficient (experimentally determined based on the rheology of the silicone), and $V_{offset}$ corresponds to the valve opening threshold voltage. Specifically, the $K_p$ coefficient was calibrated through an empirical trial-and-error procedure during the initial commissioning phase. Starting from the baseline viscosity of the silicone, the analog output was iteratively adjusted by observing the physical bead width, particularly during maximum robot acceleration and sharp corner decelerations. Furthermore, to effectively manage complex dynamics such as temperature-induced viscosity fluctuations, this calibration is integrated within a closed-loop intelligent control architecture, as suggested in recent analytical studies [15]. By means of this algorithm, when the robot decelerates during cornering, the analog voltage is automatically reduced to fully compensate for the residual volume (RV) inertia inside the nozzle. This dynamic adjustment prevents structural thinning and silicone accumulation (blobbing) at the corners [7], ensuring the bead cross-section remains perfectly stable and maintaining a standardized total silicone consumption of exactly 8.20 ml per cycle.