Dökme-taş korumalı kıyı yapıları, arkasındaki alanı şiddetli dalga ve akıntılara karşı koruyan yük-sek maliyetli yapılardır. Bu tür yapılarda, maksimum dalga tırmanması, tepe yüksekliğini belirleyen önemli bir parametredir. Kapsamlı literatür incelemesinden görülmüştür ki, regresyon analizleri neticesinde bulunan Van der Meer ve Stam yaklaşımı, dökme-taş korumalı kıyı yapılarının kret se-viyesini hesaplamak için ilgili birçok şartnamede tavsiye edilmektedir. Oysa regresyon analizi neti-cesinde bulunan bu denklemin sağlaması gereken birtakım önkabuller vardır. Bu kabullerden her-hangi biri sağlanmıyorsa regresyon analizi neticesinde elde etmiş olduğumuz denklem taraflı sonuç verir ve hatalıdır. Bu kabuller sırasıyla doğrusallık, normallik, şartlı dağılımın ortalamasının sıfır olması, eşit varyans, iç bağımlılık ve ölçülerin hatasız olmasıdır. Van der Meer ve Stam tarafından elde edilen laboratuar verilerinin istatistiki analizleri yukarıda verilen 6 şart için yapıldığında test sonuçları göstermiştir ki Van der Meer ve Stam denklemi, mevcut hali ile regresyon analizi yapılma şartlarını sağlayamamaktadır. Bu yüzden bu denklem taraflı ve hatalı sonuç vermektedir. Bu çalış-mada içerisinde hiçbir önkabul barındırmayan Bulanık Mantık yönteminden yararlanılmıştır. Nü-merik hata kriterleri ve grafik gösterimler, Bulanık Mantık yönteminin, Van der Meer ve Stam yak-laşımından daha gerçekçi ve daha pratik sonuçlar verdiğini göstermektedir. Ayrıca, Bulanık Mantık yönteminin kıyı benzerlik parametresine bağlı bir geçiş bölgesi bulunmamaktadır. Bundan dolayı, geliştirilen Bulanık Mantık modelin kullanımı kolay ve pratiktir. Bu yüzden, bu çalışmada öneril-miştir.

Extensive researches have been conducted about wave runup on smooth sloped breakwaters all over the world for a long time. Runup behavior on rough rock armored slopes are significantly different even under the same wave conditions. However, investi-gations on rock armored slopes are very limited. Rock armored slopes such as breakwaters and re-vetments are high-cost structures that defend the area behind them against severe wave attacks and strong currents. Wave runup is a crucial factor dic-tating the crest level of these structures. Van der Meer conducted extensive series of physical model tests under random sea states for investigating sta-bility of rock armored slopes at Delft Hydraulics Laboratory, during which runup values are meas-ured simultaneously. After qualitative review of var-ious parameter influences concerning wave runup on impermeable, permeable and homogenous rock structures, Van der Meer and Stam proposed the most widely used formulae for the forecast of dimen-sionless 2% wave runup elevationas as function of surf similarity parameter by using regression analy-sis, which is also recommended by the U.S. Army Corps of Engineers, British standards. The regres-sion model consists of two parts, namely linear and power functions.. Hence, it is necessary to consider a transition region which depends on the variable surf similarity parameter. Rock armored slopes are often designed according to the runup level exceed-ed by 2% of the incident waves, $R _{u2%}$ /Hs, on the front face of the coastal structure by using Van der Meer and Stam formulae. This parameter is defined as the vertical distance between the still-water level and the elevation exceeded by 2% percent of the runup values in the distribution. This means that, for every 100 waves running up a slope, two waves would have a runup elevation exceeding the level estimated by R From the literature survey, it is seen that the two most important factors influencing runup phenomena on rock armored slopes are struc-ture permeability and surf similarity parameter Since the relationships between wave runup and the-se parameters are complex, vague and uncertain in nature, it is quite difficult to adequately examine wave runup by conventional regressional approach-es. Here, an attempt is made to construct various Takagi-Sugeno (TS, 1985) fuzzy models for predict-ing the 2% wave runup on rock armored slopes. The key task in developing a satisfactory forecasting model in the TS approach is the selection of appro-priate input variables and numerical counts and types of Membership Functions of these variables, which determines the architecture of the model. Sub-sequently, the premise and consequent parameters are optimized by using ANFIS learning procedure. ANFIS, free of restrictive assumptions, optimizes premise and consequent parameters by gradient de-scent and least-squares methods, respectively, in order to best learn information about the dynamic system and does not contain any transition regions, as in the Meer and Stam equation depending on surf similarity parameter. In this study, many ANFIS models with different architectures are trained with the same 100 training data of Van der Meer and Stam. The training performance of each model is accomplished within a short time. They are com-pared with conventional empirical model of Van der Meer and Stam and with one another by using the 161 test data. As a result, the system with two inputs, namely structure permeability and surf similarity parameter, and each of which is assigned three tra-pezium MFs provides a smaller error compared to the empirical and other constructed ANFIS models and therefore is selected. In the constructed model, structure permeability and surf similarity parameter variables are initially assigned three trapezium MFs with equal base widths. Then each parameter in the membership functions in the premise part of fuzzy IF–THEN rules that changes the shapes of the mem-bership function is set to an appropriate value to match the output data of the training data by using the back propagation algorithm. This leads Struc-ture Permeability and Surf similarity parameter var-iables to different trapezium shapes. The developed fuzzy model with two inputs namely structure per-meability and surf similarity parameter yielded the best result out of all constructed models and is pro-posed in this study. The presented model is validated by comparison with widely used empirical model of Meer and Stam recommended by the U.S. Army Corps of Engineers, using the experimental data-sets of Van der Meer and Stam. The verification process is obtained through scatter diagrams and two nu-merical error criterias. It was found that the Fuzzy Logic model produce better accuracy in perfor-mance than the Van der Meer and Stam’s empirical model.