2019 年 85 巻 871 号 p. 18-00258
Our previous research (Naganuma, Hashimoto, 2018) demonstrates that the anomaly detection in the injection molding clamping unit will be made possible by evaluating the operating mode. Based on our previous research, this article suggests the sign of failure in the injection molding clamping unit can detected the anomaly detection through the unusual values of the production machine. A plastic injection molding manufacturing company built a data network system that collects a machine state log in the factory and saves data in the server. Over the past two years, two failures of clamping unit in the injection molding machine occurred under the production and the machine data of both cases are logged . We use a threshold method to detect the anomaly of the clamping unit. This thresholding is obtained by the maximum accuracy of the confusion matrix. It was tested against the two failure cases, which number of data is about 1.8 million cycle data. As a result, we discovered that this threshold method can detect the anomaly for both cases before the machine has to be shut down by the failure. Thus, it is expected to be useful for the preventive maintenance and the quality assurance of products due to its ability to detect the sign of failure in the injection molding machine.