Bulanık mantık ve gri ilişkisel yöntemleri ile bütünleşik risk değerlendirme analizi =Integrated risk assessment analysis with fuzzy logic and gray relational method

Abstract

Bir işletmede oluşabilecek hataların meydana gelmeden önce belirlenmesi, önceliklendirmesi pazarda rekabet avantajı sağlamak açısından oldukça önemli bir paya sahiptir. Bu hataların önceden tespiti ve önceliklendirmesi işletmelerin başarılı olabilmesi için kritiktir. Bu tür analizlerin yapılmaması işletmeler için maddi manevi çok büyük hasar ve kayıplara da yol açabilmekte, tahmin edilemeyen ani zararlar oluşturmaktadır.Bunların yaşanmaması adına birçok teknik geliştirilmiş olup kullanılmaktadır.Kullanılan bu teknikler ile süreçlerde iyileştirme çalışmalarına yer verilerek kalitesel, sistemsel olumlu gelişmeler yaşanmaktadır. Literatürde de hataların tespiti birçok yöntem yer almaktadır. Hata türleri ve etkileri analizi de tercih edilen yaygın bir yöntemdir.Klasik yaklaşım olan hata türleri analizi her ne kadar çok kullanım kolaylığı için tercih ediliyor olsa da bazı durumlarda dezavantajlara yol açmaktadır. Bu dezavantajlar, uzman bilgisine ihtiyaç duyuyor olması, belirsizlik ve esneklik sorunudur.Klasik yöntemin getirdiği bu olumsuzluklar bulanık metodolojinin ve gri ilişkisel analizin geleneksel HTEA tekniğiyle bütünleştirilmesiyle ortadan kaldırılmıştır.Bulanık mantık ve gri ilişkisel yöntem daha esnektir. Bu çalışmayla bir kablo üretim işletmesinde uzman bir ekiple 41 hata tespit edilmiştir.Tespit edilen hatalar kablo üretimi boyunca karşılaşılan hatalar olmaktadır. Tespit edilen her bir hata için belirlenen olasılık, şiddet, keşfedilebilirlik değerleri hata türleri ve etkileri, bulanık mantık ve gri ilişkisel analiz yöntemleriyle incelenmiştir. Önemli üç kriter olan olasılık, şiddet, keşfedilebilirlik değerleri göz önünde bulundurulmakta ve RÖS değerleri hesaplanmaktadır. Bulanık mantık yöntemi uygulması için Matlab programından yararlanılmış ve Mamdani yöntemi uygulanmıştır. Değer aralıkları üç yöntemde de birbirlerinden farklı olduğu için bu şekilde karşılaştırma yapılması sağlıklı olmamaktadır bu yüzden belirlenen RÖS değerlerine normalizasyon işlemi uygulanmıştır ve uzman görüşünün de katkısıyla her yöntemdeki 41 hata kendi içerisinde sıralanmış, yöntemler arası karşılaştırma yapılmıştır.Hesaplanan RÖS değerleri arasında istatiksel testler uygulanmıştır. İstatiksel testler için SPSS programından yararlanılmıştır. Üç yöntemin kendi aralarındaki anlamlılık düzeyleri incelenmiştir.Hata türleri ve etkileri analizinin diğer iki yöntem olan bulanık mantık ve gri ilişkisel analiz ile arasında anlamlı fark bulunduğu tespit edilmiştir. Bulanık mantık ve gri ilişkisel analiz arasında anlamlı fark tespit edilmemiştir.

Determining and prioritizing the errors that may occur in an enterprise before they occur has a very important share in terms of providing a competitive advantage in the market. There are many methods in the literature for the early detection and prioritization of these failures. Failure modes and effects analysis is also a common method of choice. There are 4 varieties. System fmea is a method applied to detect the errors of the systems or the systems under them. It aims to reduce the risk of malfunction during the operation of the system. Develops error-preventing methods by understanding and recognizing the system well. Design fmea, its purpose is to detect errors during the design phase before the production phase of the products. This is the method used to determine why the identified errors occur. During the design, customer requests are of great importance. Process error type and effects aim to eliminate the errors that may occur during the production of the product, that is, during the production and assembly stages. It tries to understand why processes experience such errors. It examines the process in general with all kinds of inputs as manpower, machine, method, material and environment. It contributes to the recovery and development of the process. Service FMEA aims to prevent these risks by determining the risks that may occur before they occur to the customer and taking necessary precautions against them. Minimizes service errors. By examining the flow, it determines and eliminates the risk of errors that may occur. It also includes maintenance and repair activities.Also FMEA has advantages and disadvantages.These advantages are there is continuous improvement and development. By increasing the communication within the team, it prepares the environment for teamwork and brainstorming, plays a role in increasing the communication in the organization. It increases the image of the companies, thus increasing their power while competing with their competitors. Considering these benefits, companies achieve high reliability. It contributes positively to their earnings. These disadvantages are the need for expert knowledge, uncertainty and flexibility.Problem arising from error types analysis has been eliminated by integrating fuzzy methodology and gray relational analysis with traditional FMEA technique. Fuzzy logic is based on thinking like a human and uses them mathematically. It is a branch that does operations by turning it into functions. Fuzzy logic, classical logic does not use. Classical logic; yes-no, 0-1, yes-no, good-bad, but fuzzy logic is this binary It also takes the values between the values for example; a little, a lot, a little, normal, medium, long. Advantages of fuzzy logic; does not need mathematical model, linear It gives good results on systems that don't have it. It is quite easy to apply, and applications are more It provides quick results. Fuzzy logic disadvantages, the method is done by trial and error so it takes a long time may be required. Fuzzy logic consists of 4 parts; fuzzification, rule base, inference mechanism, defuzzification. Fuzzification, from the outside real numeric value received from membership functions It is the process of converting them into linguistic expressions.Defuzzification, fuzzy inference of the fuzzy set transferred from the motor unit. It is the process of converting it to an exact value. The fuzzy rule base characterizing the knowledge, skills control strategy where control rules are expressed linguistically is the part. Fuzzy inference engine, fuzzy logic over rules executes and input using fuzzy rule base It establishes a connection between the output space and the output space. It information in the unit is usually given by Mamdani and Sugeno. Modeled using methods.Gray Relational Analysis; to analyze the uncertainties in multi-criteria decision problems. It is one of the methods used in mathematics and in cases where uncertainty is in question. It offers an easier solution compared to analysis methods. Classification and decision making technique. It can be applied in situations where there is a lot of data and uncertainty. It is an alternative and effective approach.In this study, 41 errors were detected in a business. Detected errors are the errors encountered during cable production.The probability, severity, and discoverability values determined for each error were examined with error types and effects, fuzzy logic and gray relational analysis methods. Probability, severity, and discoverability values, which are three important criteria, are taken into consideration and RPN values are calculated. Probability is the frequency of the error and how often a type of error can occur as a result of a particular cause. Severity, the rating of the impact of risk. Reducing the severity of the risk prevents its course from worsening. The definition of severity is the effect on the end user of the risk of error occurring. The greater the severity of the error, the greater the degree of its effect. Discoverability is a type of evaluation that exists to determine the cause of the risk. It allows the detection of potential errors and malfunctions that may occur before the product or service reaches the user. RPN stands for risk priority. It occurs by multiplying the probability, severity, and discoverability values. RPN value is not a standard value. Prioritization of errors is performed with the RPN value.The cost of continuous discoverability is high and cumbersome.Each method was listed according to the determined RPN values and expert opinion, and comparisons were made between the methods. As a result of the ranking of the RPN values, a change was observed between the rankings. When the data is examined, it is determined that some error codes have the same order for all three methods. The error type with code F31(the tube dimensions are not suitable for the technical drawing) in the 26th place among the three methods.The error with the code F16(Lack of transparency in the marking process) is at the bottom in the priority order of all methods. In terms of ranking, very close rankings were also determined for examples ; F35(No retouching of the cable),F36(Excessive sanding in retouching), F29(Not snagging or heating the macarons), F25(The macron dimensions are not suitable according to the technical document), F21(Damaged materials used), F19(The label is not suitable according to the technical document), F14(The use of marking operations on the wrong branches).Matlab program was used for fuzzy logic method application. Mamdani method was applied. The reason for using the Mamdani method is that it is a fuzzy logic method that is widely used, requires expert knowledge and can be applied to the solution of all kinds of problems.Before making comparisons between methods, data were normalized.Statistical tests were applied between the calculated RPN values. and it is desired to determine whether there is a malicious difference. Significance levels between them were examined.There is a significant difference between number 1 (FMEA) and number 2 (fuzzy logic), number 3 (grey relational analysis). There is no significant difference between number 2 and number 3. It states that fuzzy logic and gray relational analysis, which have a significant difference with FMEA, will be healthier and more successful in terms of detecting and preventing the problems that may occur in the evaluation of risks. On the other hand,there is no significant difference between gray relational analysis and fuzzy logic. Comparisons were made between the features obtained as a result of the application in terms of advantages. These comparisons have been made with limitations.In the first part of the study, a literature review was made. In the literature, studies on error types and their effects, fuzzy logic, gray relational analysis have been examined. The next section includes the definitions of these three methods, and the last section includes their application and analysis.