The research objective is to implement maintenance system model based on reliability with four proposed of
maintenance management action : combining maintenance schedule or harmonization, using the newest technology
component or inovation, repairing after hazard or resetting, and changing component which caused revitalization on the
component, for example by changing component before maintenance schedule.
The research method is application of Bathtub Failure Rate assumption with preliminary data Time to Failure
and Log Book System of Oil System in Caraka III as preliminary data, until the end of system operational in 25 years or
The result is inovation causes of retardation on reliability declining is the most effective compare the others.
Other result presents that system with or without management action in the end of operational time fulfill the requirement
of Indonesian Classification Bereau because the average of reliability for each time of system operational is bigger than
minimum requirement 0.45. Transportation Department colaborated with JICA (2002) classification proposed the
minimum average 0.88. This system cannot be reached with or without management action on the oil system reliability.
Keywords : Time to failure, Management action, Resetting, Revitalization
Dd0 transformer with a capacity of 2000 KVA mensupplay power distribution system in Timor Leste. But
as often happens more load on the transformer and the limitations of the service, the result is often damaged
Usually the damage occurs on one phase at a transformer, so it will cause a total blackout. This is due to
the limited stock available transformer there is very limited, so to overcome this, then the transformer in circuit
Open – Delta in order to keep functioning.
Transformer is connected Open – Delta then the power is supplied by only 66.7%, so the settings are in the
division of the burden on consumers, by way of rolling blackouts while waiting for the procurement of a new
Keywords: Open – Delta, Dd0 transformers, Power.
Part engine repaired process by weld always rises the distorsion at the weld result. The distorsion rise
at the weld result always due a problem that can cause increase the repaired process time and the cost of
work. The distorsion happened at weld result can form in longitudinal direction, transversal direction or
combination. The experiment limited only one direction in transversal direction. The experiment needed to
find the effect of thickness and current toward the distorsion.
This experiment used variable are thickness and current which the thickness used are 8 mm, 13 mm
dan 20 mm and the current used 125 A, 150 A and 175 A then the experiment respon used distorsion. The
design experiment developmented use central composite design (CCD) that arrangerement by software
The experiment result must be validated statistically by software minitab R14 that will found error
correlation statistically and show process model that explaining correlation between variable process
toward respon. From The matematic model show the thickness give positif contribution toward distorsion but
weld current is decline. So more thickness will be welded can increase distorsion to weld result and more
higher weld current used in weld can decrease distorsion toward weld result.
Keywords : Weld parameter, Distorsion at transversal direction, The weld result.
Improvement of education quality in university can be seen by the highness rate of success student
and the lowness rate of failed student. One indicator of failed student is drop out. The drop out problem is
interesting to be studied, because it can be affected from many factors. Many researchers study and predict
the drop out by internal factor only, which comes from the student themselves. Whereas there are many
factors besides the internal factor that can trigger drop out, such as student social behavior. However, it is a
non trivial task to determine and learn the correct classifier based on the student social behavior to predict
drop out probability.
To overcome that problem, a new model is proposed in this research to study the correct classifier
that can predict drop out using educational data mining with neural network approach.
The result from this analysis of social behavior and the proposed method is be able to show that the
use of student behavior data can increase the drop out prediction accuracy of 98,91% and the sensitivity of
social behaviour variable of 4,737.
Keywords: Drop out prediction, Educational data mining, Neural network
Feature Extraction is a significant topic in classification problem. Until now, there is no standard way
to determine best features of data. In this research, grammatical evolution with multiple fitness evaluation
approach (named as GE Multi) has been developed to extract best features of data.
The method generates a features to separate data, with n is number of classes. Some other methods
have also been evaluated in this research, including genetics algorithm, grammatical evolution with global
fitness measurement, and Gavrilis’s grammatical evolution.
It is shown in the experiment that GE Multi produces better results compared to the three other
methods using decision tree classifier.
Keywords: Feature-extraction, Grammatical evolution, Classification, Multi-fitness
- Jurnal Internasional
- Jurnal Publikasi Dosen
- Jurnal Strata 1
- Jurnal Strata 2
- Jurnal Vol 12 No. 1 Jan 2009
- Jurnal Vol 16 No 1 Mei 2012
- Jurnal Vol 16 No.2 Des 2012
- Jurnal Vol 17 No 1 Mei 2013
- Jurnal Vol 17 No 2 Des 2013
- Jurnal Vol 18 No. 1 Mei 2014
- Jurnal Vol 18 No. 2 Des 2014
- Pedoman Penulisan Jurnal
- Proceeding SNTEKPAN