Flexible electronics are becoming the next generation of devices due to their advantages, such as mechanical flexibility, eco-friendliness, large-area applicability, and scalability for mass production. However, solution-based manufacturing processes are prone to defects like discontinuities and local smudging, which can significantly degrade both device quality and yield. To tackle these challenges, rapid and accurate defect classification is crucial for real-time diagnosis during manufacturing. This study investigates the impact of data scale and key training hyperparameters on the performance of deep learning–based defect diagnosis models, using a dataset of conductive pattern defects in flexible electronics. We specifically examine how the number of training images affects model accuracy and generalization, and we analyze how adjustments to hyperparameters—such as L2 regularization and dropout—influence model performance in data-limited scenarios. Our findings offer insights into optimal training strategies tailored to different data scales and learning constraints, providing practical guidelines for designing and developing AI-based defect diagnosis models for flexible electronic devices.

This study introduces a straightforward and cost-effective method to enhance the positional accuracy of a 6-axis serial robot using a double ball-bar (DBB). Kinematic errors, a primary source of inaccuracies in offline programming, are estimated and calibrated through circular tests. The kinematics of the robot are modeled using the Denavit-Hartenberg (D-H) convention, and a mathematical relationship between radial deviation and kinematic errors is established. To avoid singularities, identifiable parameters are selected using singular value decomposition. The method involves three steps: measuring the tool center point (TCP) with the DBB, estimating key kinematic parameters, and verifying the calibration results. Redundant or less significant parameters are excluded to concentrate on the most impactful ones. During the process, the robot is commanded to trace a circular path while radial deviations are recorded. This data is then utilized to estimate and adjust the kinematic model. After recalculating and executing the circular path with the calibrated model, a notable reduction in deviation is achieved. This proposed approach requires no additional equipment and provides a quick, affordable solution for improving the accuracy of industrial robots while lowering maintenance costs.

The use of environmentally friendly, lubricant-free plastic seals in the rotating parts of robots and machines is on the rise. However, variations in seal geometry and operating conditions can influence the contact pressure between the seal and shaft, potentially leading to poor sealing performance, premature wear, or debris ingress. Therefore, advanced design optimization is essential. In this study, we conduct a parametric study and sensitivity analysis to enhance the performance of plastic seals. Finite element analysis (FEA) is carried out using a 2D axisymmetric model with interference fit contact conditions to accurately simulate the behavior of the seal and shaft. We verify the reliability of the analysis by comparing the deformation of the seal diameter before and after shaft insertion with experimental measurements obtained using a 3D tactile measurement device. We analyze four design variables: pressure, temperature, seal diameter, and coefficient of friction, considering seal contact pressure as the objective function. Sensitivity analysis is performed to determine the impact of these design variables on contact pressure and to identify trends.

Overhang structures are essential geometries in metal additive manufacturing for realizing complex shapes. However, achieving stable, support-free overhang structures requires precise control of process parameters, and securing shape fidelity becomes particularly challenging as overhang length increases due to thermal deformation. To address this challenge, this study proposed a Bayesian optimization framework for efficiently identifying optimal process parameters to fabricate high-difficulty overhang structures. An image-based scoring method was developed to quantitatively evaluate shape defects. Experimental data were collected by fabricating 3, 6, and 9 mm overhang structures with various process parameters. Based on collected data, Gaussian Process Regression (GPR) models were trained. A physics-informed soft penalty term based on energy density was incorporated to construct a surrogate model capable of making physically plausible predictions even in extrapolated regions. Using this model, Bayesian optimization was applied to overhang lengths of 12, 15, and 18 mm, for which no prior experimental data existed. Recommended parameters enabled stable, support-free fabrication of overhang structures. This study demonstrates that reliable optimization of process parameters for complex geometries can be achieved by combining minimal experimental data with physics-informed modeling, highlighting the framework’s potential extension to a wider range of geometries and processes

Microfluidic chips have become a critical component in advanced applications such as biochemical analysis, medical diagnostics, drug development, and environmental monitoring because of their ability to precisely control fluid flow at the microscale. The functionality of these chips is highly dependent on the precision and dimensional stability of microchannel structures formed on them. While injection molding is an efficient method for a mass production of microfluidic chips, it is required to minimize undesirable deformation due to thermal and mechanical stresses, which can degrade the overall performance. This study investigated global (Macro-scale) and local (Micro-scale) deformation behaviors of injection-molded microfluidic chips. Effects of processing parameters, including mold temperature, melt temperature, filling time, and packing pressure, were investigated. The Taguchi-based design of experiments approach was employed to systematically analyze these effects and to determine optimal conditions to minimize deformation.

In this study, we propose a deep learning-based method for large-area inspection aimed at the high-speed detection of micro hole diameters. Micro holes are detected and stored in large images using YOLOv8, an object detection model. A super-resolution technique utilizing ESRGAN, an adversarial neural network, is applied to images of small micro holes, enhancing them to high resolution before measuring their diameters through image processing. When comparing the diameters measured after 8x super-resolution with the results from existing inspection equipment, the average error rate is remarkably low at 0.504%. The time taken to measure an image of one micro hole is 0.470 seconds, which is ten times faster than previous inspection methods. These results can significantly contribute to high-speed measurement and quality improvement through deep learning.

Most temperature indicators that use thermocouples as sensors include an internal thermometer for compensating room temperature variations. This thermometer measures ambient temperature, which is then converted to a thermoelectric voltage. This voltage is added to the electromotive force measured in the thermocouple sensor and then converted back to temperature. Although precise calibration of the indicator can be conducted in a controlled room-temperature environment, additional uncertainty arises due to room temperature compensation during actual measurements. To address this issue, we calibrated temperature indicator at the ice point. In this experiment, the indicator was placed in an environment where the temperature varied between 8 and 38oC, demonstrating its dependency on ambient temperature. In a second set of experiments, we shorted the thermocouple input terminal to verify whether the indicator correctly indicated the ambient temperature. This study proposed a method to assess additional uncertainty that must be considered when using a thermocouple connected to an indicator calibrated with an external ice point in a laboratory. It also highlights additional steps and factors to consider during the calibration of temperature indicators that employ internal temperature compensation.

As system performance continues to improve at higher resolutions, it becomes increasingly important to establish standards for imaging degradation caused by optical windows. In this study, random surface shapes were simulated on large area optical windows with peak-to-valley (P-v) values of 0.25, 0.5, and 1.0 λ. Modulation Transfer Function (MTF) values were derived for 1,000 cases per P-v value using Monte-Carlo simulations. The specifications achieved a surface accuracy of 0.5 λ and a parallelism of 0.01 mm. MTF measurements showed that the system MTF was 13.5% prior to the installation of the optical window, and 13.1% after installation. This indicates a degradation rate of approximately 3%.

Gearboxes used in the drivetrain of intelligent robots are key mechanical components that play a significant role in determining the performance of modern robotic systems. Gearboxes employing the planetary gear mechanism, known to achieve a wide range of reduction ratios while remaining relatively cost-effective, have recently been adopted in robot drivetrains. In this paper, we utilize domestic technology to fabricate a gearbox using a compound planetary gear mechanism and conduct an evaluation of eight performance aspects of the developed gearbox through the fabrication of a dynamometer and a jig. The dynamometer comprised of the gearbox, input motor, input-output torque sensors, and a powder brake. By driving the input motor and applying braking force with the powder brake, we compare input torque sensor values with output torque sensor values to derive results. A test jig is created, consisting of an input motor, gearbox, and encoder sensor, for the measurement of inverse operation characteristics and backlash. By conducting a performance evaluation of the developed high-strength, high-reduction-ratio compact planetary gearbox, we validate the potential of the testing system and extend the scope of domestic gearbox technology development.

Among chemical coating methods, deposition using electrostatic spraying is commonly employed in coating processes to control the deposition rate, thickness, and properties of the formed materials. In this study, we considered the following variables: ring electrode, ring diameter (RD), ring voltage (RV), and nozzle-ring distance (NTR). Through experiments, we determined the atomization mode applied voltage, Sauter mean diameter (SMD), and SMD standard deviation of the nozzle. Additionally, we derived the voltage intensity and electric field along the axial direction using ANSYS maxwell to identify the optimal ring electrode atomization conditions.

Falls are common among older people. Age-related changes in toe strength and force steadiness may increase fall risk. This study aimed to evaluate the performance of a fall risk prediction model using toe strength and force steadiness data as input variables. Participants were four healthy adults (25.5±1.7 yrs). To indirectly reproduce physical conditions of older adults, an experiment was conducted by adding conditions for weight and fatigue increase. The maximal strength (MVIC) was measured for 5 s using a custom toe dynamometer. For force steadiness, toe flexion was measured for 10 s according to the target line, which was 40% of the MVIC. A one-leg-standing test was performed for 10 s with eyes-opened using a force plate. Deep learning experiments were performed with seven conditions using long short-term memory (LSTM) algorithms. Results of the deep learning model were randomly mixed and expressed through a confusion matrix. Results showed potential of the model"s fall risk prediction with force steadiness data as input variables. However, experiments were conducted on young adults. Additional experiments should be conducted on older adults to evaluate the predictive model.

Inertial measurement unit (IMU)-based 3D joint angle estimation have a wide range of important applications, among them, in gait analysis and exoskeleton robot control. Conventionally, the joint angle was determined via the estimation of 3D orientation of each body segment using 9-axis IMUs including 3-axis magnetometers. However, a magnetometer is limited by magnetic disturbance in the vicinity of the sensor, which highly affects the accuracy of the joint angle. Accordingly, this study aims to estimate the joint angle using the 6-axis IMU signals composed of a 3-axis accelerometer and a 3-axis gyroscope without a magnetometer. This paper proposes a recurrent neural network (RNN) model, which indirectly utilizes the joint kinematic constraint and thus estimates joint angles based on 6-axis IMUs without using a magnetometer signal. The performance of the proposed model was validated for a mechanical joint and human elbow joint, under magnetically disturbed environments. Experimental results showed that the proposed RNN approach outperformed the conventional approach based on a Kalman filter (KF), i.e., RNN 3.48° vs. KF 10.01° for the mechanical joint and RNN 7.39° vs. KF 21.27° for the elbow joint.

Drill processing is essential in various industries, such as automobiles and aviation. Carbide is mainly used for drilling, but cermet is also one of the most used materials. Since cermet has low reactivity with iron and low reactivity at high temperatures, excellent surface roughness can be obtained. However, experimental research comparing the performance of carbide and cermet drills is lacking. The purpose of this study was to investigate the difference in the cutting characteristics of cermet and carbide tools. The experimental conditions were feed rates of 150, 200, 250, and 300 ㎜/min and 1,000, 1,500, and 2,000 revolutions per minute. S45C was used as the workpieces. In this study, surface roughness, inner diameter, and spindle load were derived as experimental results and used as indicators to evaluate the performance of carbide and cermet drills. The results showed that the performance of the cermet drill was superior to that of the carbide drill.

Slot-die coating is a method of coating a wide layer of thin film on a substrate. It has the advantages of large-area coating with high reproducibility and uniform thickness. For this reason, it has been widely applied in various industrial manufacturing fields. To secure higher coating uniformity under various coating conditions, estimating and controlling the flow rate of the coating solution discharged to the substrate is crucial. In this study, a practical gravimetric flow rate measurement method for slot-die coating uniformity evaluation has been introduced. The gravimetric method is a technique for accurately and quickly estimating the flow rate through the mass change over time using a precision weighing balance. We analyzed the measurement principle and errors caused by fluid mechanics such as hydrodynamic force or capillary force. The dynamic properties based on fluid viscosity were also evaluated for flow rates from 5 to 50 μL/s. The repeatability of the fabricated measurement system was ~1.5 μL/s. Finally, it was confirmed that the settling time for high-viscosity fluid could be advanced by 56.4% through multi-step feedforward control.

Industrial robot manipulators require high absolute position accuracy of the end effector to perform precise and complex tasks. However, manufacturing errors cause differences between nominal and actual parameters, and errors between the expected and actual positions of the end effector, resulting in undesired lower absolute position accuracy. Accordingly, to increase the absolute position accuracy of the end effector, kinematic calibration is required to correct the nominal parameters close to the actual parameters. However, in this study, redundancy of parameters may occur from the overlapping degrees of freedom of parameters in adjacent frames, which causes the problem of unnecessarily correcting many parameters in the optimization process. Thus, to solve this problem and use only the necessary parameters, this paper focuses on the linear relationship of redundant parameters and proposes a method of automatically discriminating and removing it through the Pearson Correlation Analysis. Additionally, through simulations on the two manipulator models, we verify the accuracy of redundancy of parameters determined by the proposed method, and demonstrate consistency and efficiency by comparing the results before and after redundancy removal.