The development of a circulation control model in the drilling process using the neuro-fuzzy inference system

M70 M70 Gubitak isplake predstavlja nekontrolisano isticanje bušaćeg fluida kroz formacije kao što su kaverne, pukotine, ili drugi slojevi. Tradicionalne metode procene gubitka isplake se zasnivaju na primeni seizmičkih podataka ili pronalasku „mesta“ gubitka isplake na osnovu raspoloživih podataka iz susednih bušotina. Međutim, ove metode procene nisu pouzdane.‘ U disertaciji je izvršena analiza i procena uticaja parametara bušenja, geoloških faktora, karakteristika formacije i fluida na gubitak isplake kao i na formiranje fraktura u formacijama. Uspostavljeni su modeli koji podrazumevaju proces obrade i izbor bušaćeg fluida koji je kompatibilani sa karakteristikama izbušene formacije. Cilj disertacije je stvaranje složenog modela za analizu i predviđanje gubitka isplake tokom procesa izrade kanala bušotine, i to primenom fuzzy logike i neuronskih mreža. U radu je opisan problem gubitka isplake, njegovo predviđanje i prevencija. Objašnjena je koncepcija fuzzy logike i pravila na kojima se ona zasniva, radi boljeg razumevanja problema.Detaljno je opisana tehnika adaptivnog neuro - fuzzy sistema zaključivanja (ANFIS) uključujući jezički aspekt, numeričke informacije za integraciju kombinacije predstavljenih podataka i način na koji ovaj model može da se poveže sa gubitkaom isplake. Disertacija predlaže ANFIS model za predikciju gubitaka cirkulacije isplake tokom bušenja u naftnoj i gasnoj industriji. U ovoj disertaciji ANFIS model je upoređen sa kNN (eng. k-nearest neighbors k- najbližih suseda) modelom i metodom stablao odlučivanja (MSO). Rezultati su pokazali da se ANFIS i PSO_ANFIS (Particle Swarm Optimization_ANFIS) modeli mogu kvalitetno koristiti u slučajevima prevencije gubitka isplake i da ANFIS model obezbeđuje bolje predikcije od kNN i MSO modela. The circulation loss represents uncontrollable flow of mud into a formation which can be found in natural caverns, cracks or other layers. Traditional estimation techniques of lost circulation often use seismic data, or find the lost circulation location according to adjacent well data. However, these efforts are still not enough because the precision and details in the applied methods have not been as good as possible. In this dissertation, the effect of drilling parameters, geological factors, and characteristics of formations and fluids on the lost circulation process or formation of fractures within the layers is predicted. In this regard, some models are established which encompass the processes of treatment and choice of fluid compatible with characteristics of drilled layers. The aim of the dissertation is the creation of a complex model for analysis and prediction of the circulation lost during the drilling process, using fuzzy theory and neural networks. The dissertation describes in detail the problem of circulation loss, its prediction and prevention. In addition, the procedure of fuzzy logic and the basic rules on which it is based, for better interpretation of the problem. The adaptive neuro - fuzzy inference system (ANFIS) technique, model characteristics and how this model can be related to circulation loss models are also described in detail. The dissertation proposes an ANFIS model for the prediction of mud circulation losses during drilling in the oil and gas industry. In this dissertation, the ANFIS model is compared with the kNN model and the decision tree method. The research results in this dissertation have shown that the applied ANFIS and PSO_ANFIS (Particle Swarm Optimization_ANFIS) models can be used well in cases of mud loss prevention. The results also show that the ANFIS model provides better predictions than the kNN and MSO models.