Detection of Machines Anomaly from Log Files in Hard Disk Manufacturing Process using Decision Trees

Thanatarn Pattarakavin, Prabhas Chongstitvatana
Faculty of Engineering
Chulalongkorn University

Thai journal of operation research, vol 4, no 2 (July-Dec 2016), pp.10-17.


Hard  disk  manufacturing  is  an  important  industry  in  Thailand.    The  production  line  of  its manufacturing  process  is  highly  complex  and  consists  of  hundreds  of  automated  machines  running  a continuous flow production.  When an anomaly event occurs,the production line has to be stopped and the diagnosis engineering team must identify and locate the source of error with the machines and correct it quickly.  In an automated production line, all of the machines are monitored and their log files are sent to a server continuously;engineers then use these log files to diagnose the causes of the errors.  This work proposes to use the machine learning method to identify the anomalous events in the log files. such as an anomaly caused by engineers, anomaly from systems, and anomaly from software.  The experimental results showed that accuracy was100%using 10-Fold validation, so it is very accurate and it can help teams of engineers perform diagnosis quickly and effectively.

Keywords: hard disk manufacturing, head stack assembly, log files,machine learning,decision tree