Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : December 02, 2021 - December 03, 2021
In recent years, owing to the effects of the global coronavirus pandemic and decline of the working-age population, production efficiency and labor productivity are being actively improved in manufacturing and production sites by the introduction of digital technologies. Given these trends, we have developed a system to determine the manufacturing statuses of machines at the early stage to aid production efficiency within a manufacturing company. To detect failures and abnormalities in hydraulic presses at the early stages, we have built and commenced operation of a failure diagnostic system using machine learning. A sensor that measures vibration acceleration is attached to the cylinder, oil pump, and pump drive motor, which are the main parts of the press machine; signals are continuously collected, and the signal for normal operation is modeled on the basis of the standard deviation, crest factor, and maximum signal values. Failures and abnormalities are also detected on the basis of the amount of temporal changes and deviations of the model. Thus, the possibility to predict the time to failure by monitoring such variations is shown.
