Perpustakaan Digital ITB

Maintenance is crucial for industry sustainability as it directly affects cost, reliability, safety, and productivity. In aviation, unplanned aircraft downtime adds significant costs and creates safety risks. To predict aircraft component failures, the Weibull distribution is used because it models time-to-failure data accurately. The three-parameter Weibull distribution is the most used model in aviation maintenance, with three parameters: scale parameter (?), shape parameter (?), and location parameter (?). Previous research by the Engineering Design and Production research group at ITB created a Python-based application to estimate these parameters using R-squared method applied to Weibull Paper Plot transformations. Although the application succeeded, the results and method lack validation. Without validation, the application cannot be recommended for aircraft maintenance task selection, where inaccurate parameter estimates could lead to wrong maintenance intervals or wasted costs. The author undertook this research to achieve three objectives: to find a method that can validate the results of the application, to compare the parameter estimates from the Python application with estimates from Excel Solver, and to test the user experience of the developed application. To validate the application, the author created synthetic datasets from known three-parameter Weibull distributions representing three failure modes: wear-out failures (? > 1), random failures (? = 1), and infant mortality failures (? < 1). Datasets with different sample sizes were generated from histograms with 15 classes. Both the Python application and Excel Solver were applied using R-squared. Results show that both methods produce equivalent parameter estimates with location parameter errors typically less than 5%. The Python application can estimate parameters for ? ? 1 cases while Excel Solver cannot converge in these regions. Other than that, RMSE is also calculated using Excel Solver which has the same result as R-squared method. The author concludes that both methods are valid for three-parameter Weibull parameter estimation, with the Python application offering better numerical stability across all failure rate behaviours relevant to aircraft components.