An artificial neural network model based on a deep learning algorithm is known to be more accurate than humans in image classification, but there is still a limit in the sense that there needs to be a lot of training data that can be called big data. Therefore, various techniques are being studied to build an artificial neural network model with high precision, even with small data. The transfer learning technique is assessed as an excellent alternative. As a result, the purpose of this study is to develop an artificial neural network system that can classify burr images of light guide plate products with 99% accuracy using transfer learning technique. Specifically, for the light guide plate product, 150 images of the normal product and the burr were taken at various angles, heights, positions, etc., respectively. Then, after the preprocessing of images such as thresholding and image augmentation, for a total of 3,300 images were generated. 2,970 images were separated for training, while the remaining 330 images were separated for model accuracy testing. For the transfer learning, a base model was developed using the NASNet-Large model that pre-trained 14 million ImageNet data. According to the final model accuracy test, the 99% accuracy in the image classification for training and test images was confirmed. Consequently, based on the results of this study, it is expected to help develop an integrated AI production management system by training not only the burr but also various defective images.

Plastic injection molds used for rapid heating and cooling must minimize surface damage due to friction and maintain excellent thermal and low electrical conductivity. Accordingly, various surface treatments are being applied. The properties of Al₂O₃ coating and DLC coating were compared to find the optimal surface treatment method. Al₂O₃ coating was deposited by thermal spray method. DLC films were deposited by sputtering process in room temperature and high temperature PECVD (Plasma enhanced chemical vapor deposition) process in 723 K temperature. For the evaluation of physical properties, the electrical and thermal conductivity including surface hardness, adhesion and wear resistance were analyzed. The electrical resistance of the all coated samples was showed insulation properties of 24 MΩ/sq or more. Especially, the friction coefficient of high temp. DLC coating was the lowest at 0.134.

In plate forming technology, cylindrical drawing process is widely used in industry due to technological development. In this study, we used stainless steel 3042B and stainless steel 304J1, which are the most commonly used materials in the production of cold and hot water tanks for water purifiers, among cylindrical drawing products. Under the same conditions, the thickness of the sidewall of the product formed by drawn experiment was studied. As a result of the experiment, the bottom thickness of stainless steel 304J1 was considered to be thick. It is judged that the defect rate can be reduced by changing the breaking phenomenon of the floor surface of the cold and hot water bottles to the material of stainless steel 304j1. Stainless steel 304 2B material shows a sharp change in thickness from punch corner R to sidewall position, while stainless steel 304J1 material showed a uniform change from punch corner R to sidewall position. Stainless steel 304J1 material is considered to improve the clamping of the product in the process of extracting the product after hand drawing. The appearance of stainless steel 3042B products is considered to produce more wrinkles in the flange, which exerts greater tensile force on the sidewall during molding, resulting in uneven sidewall thickness.

The bending process of a press die is to bend a flat blank to the required angle. There are V-bending, U-bending, Z-bending, O-bending etc. for bending processing, and the basic principle of calculating the unfolding length of die processing is used as the neutral plane length. Since the constant of the length value of the neutral surface is different depending on the type of bending, it is impossible to accurately calculate it. In particular, Z-bending processing is performed twice, and it is set on the upper and lower surfaces of the blank, and bending processing occurs at the same time as the upward and downward bending, and the elongation of the material occurs and the material increases. It is not possible to check with the calculated value, and it occurs in many cases where the mold is modified after start-up. This study aims to minimize die modification by developing a formula to calculate the development length of Z-bend.

In this study, laser-induced periodic surface structure (LIPSS) was fabricated on Ni, Si, and GaAs samples using a flat-top beam with a uniform energy distribution that was fabricated using a Gaussian femtosecond laser with a mechanical slit and tube lens. Unlike the Gaussian beam, the flat-top beam has a uniform beam profile, therefore the center and the periphery of the fabricated LIPSS have similar line periodicity. In addition, LIPSS was obtained not only in metals but also in metalloids and metals and metalloid compounds by using the narrow pulse width characteristic of a femtosecond laser.

Plate-shaped works are one of the materials that can be applied to the entire industry due to their various shapes and sizes. Plate-shaped parts workpieces are thin and wide, and when processing is completed, they are often bent or deformed in various directions, making it difficult to produce normal products. In particular, this study intends to study the processing deformation and distortion of plate-shaped parts fastened to the jig during milling processing. In this study, a method for preventing deformation occurring in plate-shaped parts was derived through jig element change and CAE analysis, and this was applied to actual processing to produce products with stable dimensions. Through a finite element analysis experiment, it was found that installing two supports on the back of the plate-shaped part results in minimal deformation and the optimal distance between the two supports is 150 mm. Through this experiment, when processing a thin plate product, a support was installed in a direction opposite to the cutting force applied to the thin plate to prevent deformation of the product, thereby improving defects.

In molding of continuous fiber-reinforced thermoplastic composites, it is very difficult to impregnate between the reinforcements and the matrix since the matrix has a high melting temperature and high viscosity. Therefore, most of composite molding processes are divided in the manufacturing processes of intermediate materials called prepreg and the forming of products from intermediate materials. The divided process requires additional facilities and thermoforming, and they increase the cycle time and cost of composite products. These problems can be resolved by combining the continuous fiber-reinforced composite molding process with injection molding. However, when a composite material is manufactured by inserting woven fabric into the injection mold, poor impregnation occurs on the surface of the molded product. It affects the properties of the composites. In this paper, through an impregnation experiment using cores with different heat transfer rates and pore densities, the reason for the poor impregnation was confirmed, and molding experiments were conducted to produce composite with improved surface impregnation by inserting the mesh. And also, the surface impregnation and deformation of composites molded using different types of mesh were compared with each other.

In this study, during the injection molding process, a device was manufactured and evaluated that calculates a cooling time by measuring a vibration signal generated from a mold using an acceleration. The last two parts, one of which has a large magnitude change in the measured vibration signal of a mold, were divided into a cooling start section (paking end section) and a mold opening section, and the time difference at the relevant points was calculated as the cooling time. The cooling time was monitored on a 5-inch light guide plate mold by applying the method. The manufactured device was attached to a fixed base of mold to measure the cooling time, and data was obtained remotely using Bluetooth technology. Then, the measured cooling time was compared with the cooling time set in the injection molding machine to evaluate the accuracy. As a result of the experiment, the cooling times measured by the devices were 15.675±0.024 sec, 20.637±0.014 sec and 25.623±0.079 sec of each conditions. Also, the measurement results were shown with errors of 0.655±0.044 sec, 0.637±0.014 sec, and 0.662±0.013 sec, respectively.

Solar power generation, which is one of the methods of using solar energy, has a high possibility of practical implementation compared to other renewable energy power generation, and it has the characteristic that it can generate as much power as needed in necessary places. In addition, maintenance is easy, unmanned operation is possible, and power management can be performed more efficiently if operated in a hybrid method with existing electric energy. Therefore, in this study, numerical analysis using a computer program was performed to analyze the efficient operation and performance improvement of solar energy of the rotating condensing type solar LED street lamp. As a result, the two-axis tracking type could obtain 15.23 % more electricity per year than the fixed type, and additional auxiliary power generation was required for the fixed type by 19 % per year than the tracking type. As a result of computational fluid dynamics(CFD) simulation for PV module surface temperature prediction, the The surface temperature of the Photovoltaics(PV) module incident surface was predicted to be about 10℃ higher than that of the fixed type.

The increase in automation facilities in the injection molding industry is a very important process control item. The most important item when constructing an unmanned machine using a take-out robot is the "mold opening stroke" of the mold. The injection molding machine control method is divided into hydraulic type and electric type, and there have been few studies on the mold opening distance according to the control method. In this study, the correlation was confirmed by increasing the injection speed to 20, 50, 80, and 100% for the three types of hydraulic control method, open loop and close loop, and electric control method. Through the experiment, the following results were obtained. (1) It can be seen that the reproducibility is excellent with the electric, close loop, and open loop control methods. (2) When the injection speed is set to 50%, the mold opening distance is 263.10~263.27 mm, which is the most reproducible. (3) As a result of ANOVA, both injection speed and mold opening distance showed a significant difference in the hydraulic control method (p<0.05), but it was verified through experiments that there was no significant difference in the electric control method. Based on these results, when electric control is selected rather than hydraulic control, the reproducibility of the mold opening distance is excellent, so it is thought that the taking-out robot can take the object out of the mold more safely.