Design of expert system for train operational feasibility with Tsukamoto fuzzy inference system

Train is one of the most favorite mass transportation in the world. In 2017 train in Indonesia can bring about 341.605 passengers to many destinations. PT. Kereta Api Indonesia is a state owned enterprise that has responsibility to make sure that train is safe and works well. As we know that a train has several components to check. It is very difficult to identify whether a train is in a good condition or needs repairing. The purpose of this work is to proposes a model of operational feasibility by several main criteria: bogie, breaking system, boffer, electric coupler, and safety kit. In its experiment phase, this model uses Tsukamoto Fuzzy Inference System to decide that a train is in good condition or need repairing. In evaluation phase we compare this model with traditional method and this model shows exactly 99% of the same result. It is suggested for further work to include several methods such as Tsugeno and Mamdani fuzzy inference system.


Introduction
Currently, development of technology demands rapidly to apply technology in the environment with the maximum. In addition to helping, technology applications also make it easier for users. Therefore, the technology becomes appropriate technology [1,2].
Therefore, there are many people's needs for an information technology-based system, including for decision makers requiring technology involvement [3], such as in the field of railway transportation in Indonesia in determining the feasibility of railway operational facilities.
The determination of the feasibility of the operation of the train means is the result of a series of inspections of train components that must meet the requirements of a roadworthy threshold, to ensure the train to be used or operated in good technical condition to safeguard the safety and comfort of its users. The purpose of this work is to proposed a model of operational feasibility using Tsukamoto Fuzzy Inference System to decide that a train is in good condition or need repairing [4][5][6]. Tsukamoto fuzzy inference system mainly used in a lot of filed and show that this method is effective and accurate [6][7][8].

Data Collecting
The data attributes required by the system of determining the feasibility of the train facility with the fuzzy tsukamoto method obtained from a train depot in Indonesia as follows: bogie, boffer, breaking system, electric coupler, and safety tool kit. The source of data is a train inspector of each criterion is obtained by standard assessment of specification of each criterion. The standard criteria assessment as following table.    In observation phase, we obtained formed fuzzy members and membership function is through fuzzification stage. Then the next step by forming the rules IF-THEN and the next stage of the machine inference method tsukamoto in order to be able to change the input into output using the rules that have been determined. And the final stage of converting the output obtained from the inference stage becomes a firm value into an appropriate membership function at fuzzyfication.

Fuzzification
The input stage contains the fuzzyfication process that is the determination of the variables that affect the feasibility of the train facilities. There are 5 attributes of train station eligibility obtained from interviews and case studies at Depo Kereta Cirebon. The feasibility determination variable from the train scoring assessment data is made into fuzzy sets. For example, a K1-01006 first class train with following condition: bogie K5-TB398 = 80; breaking system = 85; boffer= 85; electric coupler = 85; safety toolkit = 90.

Bogie
Bogie variable condition with x = 80 µJelek

Rule Establishment
After determining the variables and fuzzy set, then proceed the rule-making process in the form of IF-THEN. Rule is a knowledge structure that connects data information that has been known to other information so that it can be concluded. The rule structure logically links one or more antaseden (also called the premise) located within the IF section with one or more consequences (also called conclusions) located within the THEN section. Generally

Tsukamoto Inference System
Inference is the process of converting a fuzzy input into a fuzzy output (crisp) by following the established rules (IF-THEN Rule). This stage determines the value of α and z of each rule that has been established. 1.
[R1] IF bogie is good AND breaking system is good AND boffer is good AND electric coupler is good AND safety toolkit is good THEN layak. 2.
[R2] IF bogie has damaged AND breaking system is good AND boffer is good AND electric coupler is good AND peraltan kesalamatan is good THEN unfeasible  3.
[R3] IF bogie has damaged AND breaking system has damaged AND boffer is good AND electric coupler is good AND safety toolkit is good THEN unfeasible. Tabel 6 describ summary rule in order to determine α dan z value including R4, R5, dan R6 that not explain in this article.

Defuzzification
The defuzzyfication stage is the last stage of the fuzzy logic tsukamoto technique. Changing the fuzzy output of the inference becomes a firm value using the membership function that corresponds to the fuzzyfication. This stage is to determine the final z using the average value. The way the calculation is to divide the number of α * z with the number of α (alpha). The final z value by dividing the sum of α * z by the sum of α (alpha) is 85. Then the train K1-01006 from the examination result by the assessment of each criterion (train component) using the tsukamoto fuzzy method calculation yields the final z value 85 that the train is worth operate.

Result and Discussion
Performance test of decision support system to determine the feasibility of this train facility, is done by comparing the result done in the application with the calculation result which is done conventionally. According to evaluation list, we found that this proposed model has exactly 99% accuration value. It means that Tsukamoto inference system works effectively [9] in this model [10,7,11,12].

Conclusion
Tsukamoto inference system is an effective method to solve this complex model. This method has a high number of accuration with 99%. In further work, we suggest to implement others inference system such as Mamdani and sugeno. The determine factor should be improve in order to convince and satisfy the stake holder.