SMT Machine Log File PDE Features Extraction and Analysis

Tan, Wei Ning (2019) SMT Machine Log File PDE Features Extraction and Analysis. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text TanWeiNing.pdf Restricted to Registered users only Download (2MB)

Abstract

Recently, Surface Mount Technology has been widely used to mass produce printed circuit board and it can be said as a mature technology through many years of consummation. However, there is still a common issue that has been facing in every SMT based production line which is the part drop. Part drop is a very serious case that happens in SMT production line and can be caused by many factors such as the nozzle defect, the leakage in the vacuum suction tube, etc. By doing feature extraction and analysing the data to detect the possible error happen, it is believed that this issue can be minimized, and the root cause of the part drop to happen may be determined. By using this method, the root cause of part drop can be determined in a very short period of time and the engineer or even the operator is able to complete their job with a more efficient and effective way. In this project, the features such as the details of the pickup miss or also called part drop event (PDE) happen in the address is extracted out to analyse the possible error happen in the machine. The logfile is first being located and then arranged properly to be imported into MATLAB. The features will then be given some weight and the threshold, and the maximum threshold will be defined to divide the data into three stages which are normal stage, predictive maintenance stage and maintenance stage. When the data is falling beyond the maximum threshold, it will then be located in the maintenance stage which means the maintenance must be carry out as soon as possible. If the machine is normal and running without error, it is said to be a normal stage. Predictive maintenance stage means that the machine may have some error, but the error is not that serious, and the more serious error is predicted to be happen in the machine, so the checking could be carrying out by the operator leader instead of the engineer. An analysis report will be generated to show the nozzle behaviour and the actual location of the particular error part will be shown by generating a bar graph. Thus, the engineer will not need to spend his/ her time on finding out the problem and simply make his work more efficient and effective. Thus, by doing features extraction on the PDE and analysing the data can make the work of the engineer or even the operator a lot more efficient than before.