Open AccessAnalysis on current flow style for vehicle alternator fault prediction
Author(s) -
M. A. Halim,
M. T. A. Rahman,
Norasmadi Abdul Rahim,
Abdur Rahman,
A F A Hamid,
N.A.M. Amin
Publication year - 2019
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/670/1/012042
Subject(s) - normality test , alternator , statistics , kurtosis , normality , population , fault (geology) , parametric statistics , mathematics , nonparametric statistics , sample size determination , statistical hypothesis testing , engineering , computer science , power (physics) , physics , seismology , geology , demography , quantum mechanics , sociology
Vehicle alternator is only seen as fragment piece in vehicle. This project will analyse the vehicle alternator current output flow style. A study on charging rate onto battery can be made based on this analyst. From this a prediction can be made on the vehicle alternator health and may prevent it from affecting other charging system component. Features extracted from the raw sample data are root mean square (RMS), waveform length (WL) and autoregressive (AR). These features will then go through normality test to find the sample is normally distributed or not. The normality test used in this experiment is Jarque-Bera (JB) test. After go through the normality test, it shows that need to continue with non-parametric test. Because JB test shows that p-value is less than 0.05 confidence level. Kruskal-Wallis is used as non-parametric data validation. In this test, the hypothesized value is the value is the median instead of the mean as in Analysis if Variance (ANOVA). The Kruskal-Wallis test evaluates for any significance difference in the population medians on a dependent variable across all levels of a factor. For classification K-Nearest Neighbour (KNN) is used to find the number of K to differentiate between classes. After that the K value is use in Holdout method for training and testing. Final result shows that the accuracy of this machine learning tools is 94%. This number is a good percentage to be able to called as vehicle alternator fault prediction.