Condition Based Maintenance with Logical Analysis of Data

点击翻译按钮获取中文摘要

In September 2006, the viaduct de la Concorde in Montreal collapsed. The consequences were two crushed cars, five dead and six injured. The inquiry Commission declared that this tragedy cannot be attributed to one cause only, and that multiple faults resulting from sequential causes have accumulated and resulted in these consequences. In August 2007, part of the underground structure of downtown Montreal was evacuated after spotting some indication of damage done by through-flow water. In 2010, the oil spill of British Petroleum in the Gulf of Mexico took place. The accident involved a loss of control over the pressure in the well followed by the failure of the well's blowout preventer. Eleven people died as a result of the accident and others were injured. New and old challenges continue to plague the researchers in the field of Physical Asset Health Management (PAHM), and in particular in the field of Condition Based Maintenance. The degradation and failure problems as well as some of the phenomena that physical assets are experiencing are complex. By complex, we mean dynamic and chaotic behavior due to non-homogeneities in tasks and usage conditions, noisy and non-linear, multiple interacting causes. Therefore, forecasting can be practically impossible when based on the traditional model-based predictive algorithms. Traditional as well as new techniques should be adapted to the present needs of better follow-up and maintenance of these physical assets. Logical Analysis of data (LAD) is a data mining, artificial intelligence, machine learning approach that is based on pattern recognition. Initially it started in 1988 at Rutgers University Center for Operations Research (RUTCOR), and it was applied mainly in the medical field. In 2007, it was applied for the first time at École Polytechnique by Dr. Yacout's research group in the field of PAHM. After thirteen publications, this approach seems to offer new ways for solving the above-mentioned problems. This proposal aims to continue the in-depth study, adaptation, and development of techniques based on this approach, for the field of PAHM, and the development of an integrated tele-maintenance intelligent system based on LAD.