-LSmtron leads joint study with Roh Joon-Suk research team, Professor of Mechanical Engineering and Chemical Engineering at POSTECH, Korea Institute of Production and Technology, VM Tech and POSCO.

-System that recommends injection molding process conditions by combining artificial neural network and random search is developed.

▲ WIZ Plus Injection Molding Machine

Recently, a research team led by Roh Joon-suk, a professor of mechanical engineering and chemical engineering at POSTECH, developed a system that recommends injection molding process conditions by combining artificial neural network and random search.

The research, which was conducted jointly with LS Mtron, the Korea Institute of Production and Technology, VM Tech and POSCO, and supported by the Ministry of Science and ICT, the Korea Research Foundation, the Ministry of Trade, Industry and Energy and the Korea Institute of Industrial Technology Evaluation, was published in the specialized journal "Advanced Intelligent Systems."

The research team conducted a study to find the process conditions that satisfy the desired quality after studying the relationship between process conditions and final products with artificial intelligence. First, 3,600 simulation data and 476 experimental data were obtained from 36 different molds, and as a result, each data confirmed that 15 shapes and 5 processes were input values and that the weight of the final product was taken as output values.

Based on the weight prediction model learned by introducing transference learning, a recommendation system was created to find optimal process conditions by random searching, and finally a graphic user interface (GUI) was developed to be used for actual injection machines.

As a result of verifying the process conditions recommended by the artificial intelligence model, the average relative error of 0.66% was achieved. The research team said that non-experts with injection molding can also set up process conditions that have errors within 1% of the desired weight of the product by entering shape information for any product based on the system.

The study collected information on the results (product weight) by changing both quantified shapes and process conditions for products with 36 different shapes. In other words, even if a new product is molded, the process conditions can be controlled without having to predict the results and generate learning data by simply entering the shape of the product.

By using the artificial neural network system, which is considered the biggest advantage of being able to obtain various shapes in real time, you can achieve uniform results by simply entering the shape of the product and the weight of the final product you want although you are not an injection expert,

As a result, related industries are anticipating that this achievement will enable the use of "unmanned smart factories" in various manufacturing industries such as plastic injection processes, cutting, 3D printers, and casting.

▲ Roh Joon-Suk

■ Inquiry: LSmtron Co., Ltd . http://lsinjection.com

   Edit : HANDLER  http://www.ihandler.co.kr