| Popis: |
A new automatic fault diagnostic method based on variable trend analysis was developed. Every time a fault is detected, the proposed method uses a local model to determine the variable trends associated to the detected fault. The fault identification is achieved by comparing the extracted variable trends with a previously generated library of faults signatures. The method was implemented and tested with a set of faults purposedly generated in an oil production well equipped with a pressure cavity pump (PCP). It was able to correctly identify all the generated faults, outperforming simpler versions of the method based on global models. |
