Pustaka Ilmiah Universitas Padjadjaran

Abstrak

Analysis Of Accuracy Multivariate Control Chart T2 Hotelling Free Distribution With Outlier Removal (Case Study: Production Process Ink at PT. EPSON Batam)

The quality of products are competitive advantage for PT EPSON Batam-Indonesia. One of their products is ink, its quality are controlled by collecting data during 1st April 2014 till 7th July 2014 and use these data to set up a control chart on five variables, namely, Viscosity / viscosity (DPAs), pH, Surface Tension , Wavelength (Å), and Particle Size (m). We noted that 14 points plot above the upper control limit, so the process is not in control.These points must be investigated to see whether an assignable cause can be determined. However, analysis of the data (chamber process) does not produce any reasonable or logical assignable cause for these, and we decide to retain the point and conclude that the process is in control. This indicates that we have false alarm, so clearly this is an important issue to consider in control chart implementation. The ink quality is determined by five variables which are multivariate in nature, while PT EPSON Batam is applying univariate control charts to each individual variable. Control chart is basically used on data having normal distribution, but in fact the data we have is not normal distribution so Chebyshev theorem for free T2Hotelling control chart distribution can be applied. In addition, to address the presence of outliers, the control chart is constructed without outliers (outlier data removal). The Accuracy of control chart is evaluated by G-index measurement accuracy. The proposed control chart produced 4 out of control observations and after decomposition analysis are known that the cause of out of control occurrence are the variable of the surface tension and viscosity. In addition, the accuracy of the free distribution T2Hotelling control chart, 97.94% shows that the control chart can be used for future samples.