DOI: 10.36347/sjet

Pages: 563-566

One of the major problems faced by the oilfields in the middle time is casing damage. The neural network, the improvement BP as well as the three-layer feed forward neural network applied in damage diagnosis is a key way to prevent damage. The formations of the damage are corresponded to the input vector. Factors such as casing age limit, strata stress, soaking time are described as real Numbers between 0 ~ 1. Multi -factor constitute a fuzzy vector. Thus, working status such as the output of neural network, a large number of field data and categorical reduction constitute the training simple. After those trainings, predictions of the casing status can be completed.