Muhammad Ilham Ayyasy, 

Universitas  Jember 

Skriptyan Noor Hidayatullah Syuhri, 

Universitas  Jember

Ahmad Syuhri

Universitas  Jember 

ABSTRACT 

The accuracy of Remaining Useful Life (RUL) estimation is a crucial component in optimizing the determination of condition-based maintenance (CBM) strategies. However, the prediction results of RUL models are often affected by the phenomenon of incomplete historical data in the industry, especially for older assets. This study investigates the quantitative impact of two incomplete historical data phenomena, namely the loss of early historical data that will affect the determination of the start date and the phenomenon of missing failure data that can reduce the number of failure sequences, on the performance of the 2-parameter Weibull and Crow-AMSAA reliability models. This study was conducted using an experimental approach based on data simulation, with the main data set that also serves as a benchmark that is varied using scenarios of early historical data truncation (loss early history) and data deletion (missing failure data) with simple random sampling for the determination of the start date and data elimination scenarios. The modeling results show that the phenomena of loss early history and missing failure data can distort the values ​​of the shape parameters and scale parameters that cause deviations in the results of the Weibull reliability curve and the Crow-AMSAA cumulative failure. However, the Weibull Distribution model is proven to be more robust to start date uncertainty in the intact data scenario (loss 0% early history).

Keywords: Weibull Distribution, Crow-AMSAA, Reliability, Missing Failure Data, Loss Early History