In the casting industry, high-frequency induction melting furnaces have a great impact on increasing productivity and improving the manufacturing environment, but from an industrial accident management perspective, industrial accidents are frequently ...
In the casting industry, high-frequency induction melting furnaces have a great impact on increasing productivity and improving the manufacturing environment, but from an industrial accident management perspective, industrial accidents are frequently caused by fires, explosions and ruptures caused by foundry furnaces. In particular, small and medium-sized casting manufacturing plants are always exposed to risks, which necessitates countermeasures.
Currently, foreign production equipment manufacturers of high frequency induction melting furnaces in the U.S. and Germany want to predict the remaining life of the production equipment and prevent large industrial accidents by detecting the conditions of the production equipment in advance. And by connecting IoT sensors to detect abnormal signals and By analyzing maintenance cycles of production equipment and abnormal detection signals, intelligent research activities such as AI predictive learning models are taking place.
Therefore, if the abnormal detection prediction maintenance monitoring system of this study is applied to the manufacturing plant, it will reduce the risk of industrial accidents in the casting industry and have the substitution effect of commercial S/W localization.
In this paper implemented the high frequency induced melting furnace abnormal detection and prediction maintenance monitoring system using MLP model. To collect data on the condition of production equipment operation in real time by analyzing the electrical and mechanical characteristics of high-frequency melting furnaces, It were implemented production equipment interlocking interface function that collects digital converter devices from IoT sensors and PLC modules in a stable and accurate method.
Although it depends on production operation rate, a service that collects about 200,000 monthly production equipment status data is operational and monitored. The PLC Repository Master server is established independently for reliable transmission and storage of the collected data, and the data can be analyzed and distributed processing using the abnormal detection monitoring and predictive learning model of the melting furnace production equipment. The pre-processing phase of equipment operation data collected for more than a few months was carried out to analyze the correlation of 18 data related to the fault prediction of equipment. Of these, six input data were finally selected and the predicted accuracy results were obtained 91% in 10 times learning testing using the predictive learning model.
If time series data is secured through continuous storage and monitoring management of systematically collected production equipment history data, it is believed that the learning model of time series analysis can be studied and developed by upgrading to LSTM predictive learning model in the future.