METHOD FOR REAL-TIME LITHOLOGY IDENTIFYING WHILE DRILLING

It is very important for the drilling engineer and geologist to real-timely identify formation lithology while drilling. Based on the mud logging data to identify lithology, the effect of drilling operational parameters, hydraulic factors and bit wear should be comprehensively considered. It is not practical to real-timely measure bit wear while drilling. According to BP neural network theory, dual neural network model for lithology identification was established. The first neural network was used to identify lithology under the condition of known bit wear. The second neural network was used to predict bit wear under the condition of given lithology. The two neural networks were linked by the state of bit wear. Sample data were selected to train the two neural networks respectively. Combined with mud logging data, lithology was predicted according to the process of lithology identification. This model was verified in xin jiang oilfield. Compared with geological explanation of logging data, the prediction result is better and the coincidence rate can reach about 85%. Field application result shows that this model is effective and feasible.