Gemilerin dar sulardan geçişleri, maruz kalınan bozucular ve coğrafi yapı nedeniyle çok gelişmiş seyir cihazlarına rağmen sorun olmaya devam etmektedir. Birçok ülkede dar sularda emniyetli geçişin sağlanması amacıyla, belirli aralıklarla yerleştirilmiş radarlar vasıtasıyla geçiş yapan gemilerin izlenmesi ve bu sayede çarpışma, karaya oturma ve diğer şekillerde oluşabilecek kaza risklerinin azaltılması için Gemi Trafik Hizmetleri Merkezleri (GTH) görev yapmaktadır. Bu çalışmada, İstanbul Boğazı model alınarak, GTH merkezinde erken uyarı sistemi ve geçiş yapan gemilerde seyir yardımcısı kılavuz olarak görev yapacak bir yöntem geliştirilmiştir. Boğaz kazalarının, gemilerin trafik ayrım düzeni sınırlarını ihlal etmeleri nedeniyle meydana geldiği bilindiğinden, ihlalin önceden tespiti ile tehlike riskini azaltmak ve olası kazaları engellemek hedeflenmiştir. Özellikle görüşün azaldığı, alt akıntının arttığı durumlarda ve 45 dereceden daha yüksek manevralarda kaza riski artmaktadır. GTH merkezinin imkânlarından faydalanılarak, manüel olarak kumanda edilmek suretiyle boğaz geçişi yaptırılan gemilerin tüm çevre şartlarına bağlı olan hareketine ait verilerinden faydalanarak yapay sinir ağı eğitilmiş ve eğitilmiş yapay sinir ağı (EYSA) ile İstanbul Boğazı’nın coğrafi ve oşinografik özelliği göz önüne alınarak kabul edilmiş süre olan 3 dakika sonraki yerleri öngörülmüştür. Gerçek zamanlı olarak yapılan öngörü GTH personeli ile değerlendirilmiş ve en az 2 dakikalık öngörünün kendileri için faydalı olacağı tespit edilmiştir. Vektörel hata analizleri de yapılarak öngörünün istenen seviyede olduğu görülmüştür.

Despite modern navigation devices, there are still some problems for navigating of vessels in narrow waterways because of geographical structures and various disturbances. In this study, guidance and a early warning method have been developed for navigation in narrow waterways. The Istanbul Strait has been specifically studied as a model. The basis of this study is to predict the future coordinates of a manually controlled vessel using Artificial Neural Networks (ANN). The method could be utilized for warning system by Vessel Traffic Services (VTS) operators and guidance system by vessel crew. Even though all vessels have modern navigation devices, use of autopilot is not allowed in the Strait. Vessels have to be manually controlled according to rules. A human operator (helmsman) controls a vessel as a continuous system manually according to captain’s command against random disturbances through strait of Istanbul and their behaviour affects the course and coordinates of the vessel. In this study, Artificial Neural Networks have been trained by using position and speed data collected from vessels which navigated manually in the strait and they included effects of environmental conditions and geographical characteristics of the strait. Three-minute-ahead-position of vessels have been predicted by using the trained ANN. Some experiments have been realized in Ä°stanbul VTS centre and it has been observed that the method satisfied the goal in especially turning points of the strait. VTS operators can watch only straight bearing of vessels on VTS panels but especially for turning regions, they have to notice a risk on time which may result with a disaster because of negligence of the vessel crew. Latitude and longitude (x,y), course (ψ), speed (u) and time ( t) data have been collected from GPS of 4 vessels in 1 minute sampling period to train Artificial Neural Networks. Then the trained ANN has been used to predict xˆ (t +1), xˆ (t + 2) , xˆ (t + 3) , yˆ (t +1) , yˆ (t + 2) , yˆ (t + 3) positions. Data from vessels, navigated to south and north, have been used to determine model structures. Although the course is stable on the reference line between two turning points, vessels can change its course if it is between traffic separation lines. The direction of vector, from x[k-1], y[k-1] to x[k], y[k], gives the course of the vessel. These data includes speed data because they were recorded in 1 minute sampling period. So, only position data considered to decide model structure and used for training of Neural Networks. On-line prediction has been tested at the VTS centre in Istanbul. and, it has been confirmed that, there is no early warning and guidance system as intended in this study, either in Turkey or in another country to use along with VTS systems. Warning mission is currently realised by certificated and experienced operators. The prediction results evaluated together with VTS operators. They found the prediction performance satisfactory and stated that they need minimum of 2 minutes to prevent a probable risk. Currently a Straight bearing line is shown on the VTS operator panels and it represents vessels’ course but it’s not useful especially on the turning regions. The proposed method has been particularly formed for prediction of turning course and the prediction performance was found to be successful on the turning regions. Error analysis has been made for examining differences between predicted position and real position of the vessel. Three different methods can be suggested for prediction. As in this study; data file of the vessel has been updated and Neural Networks have been trained in 1 minute sampling period and then 3-minute-ahead position has been predicted with trained Neural Network. As another possible method; Trained Artificial Neural Networks (TANN) libraries can be constituted by collecting data statistically from vessels navigate in Strait depending on the current environmental conditions. Finally as a third option, special TANN libraries can be constituted by keeping the past data of any specific vessel and it can be later used for the same vessel and of course environmental conditions should also be considered. In addition, if VTS operators can watch the course prediction instead of straight bearing on operator panels for vessels navigating in opposite directions in straits, the algorithm can be expanded to make “collision prediction” for early warning in especially turning regions and this issue can be studied also in future.