Application Research Of Bayesian Networks Into Mechanical Fault Diagnose

With the improvement of mechanical on automatic, intelligence, large scale, integration and complication, it is the very turn to start out diagnose of mechanical failure, which is based on the check-up of function condition of machine. To carry out some relative of function as well as we can ensure the reliability of modern machine diagnose by means of information process, the potential combination of many failures that happen at the same time, as well as the shortcomings of unanalysable, the author will focus more on the Bayesian Network.The Bayesian Network is one of the most efficient theoretically model on uncertain things expression and inference field. If can be used to the expression as well as inference of uncertainty and probability. As a exist-way diagram description based on net construction, it can start two-way inference as well as synthesize pre-tested and sample information. The result seems conceivable of we follow that. Since that, Bayesian Network has a deep impact on mechanical failure field.With the target of some common vibration failure in rotary machine, the author gives a depiction of some features firstly. Then elaborates the basic theory on Bayesian Network, the accurate reference approaches based on junction tree, and some learning methods. We set up some mechanical failure models. Features as follow:(1) Bayesian Net model can melt into the uncertain things of mechanical failures diagnose naturally, and also have a good foundation of probability theory to make it more reasonable and accurate.(2) The expression ways of diagram-like is much clearer, quite understandable.(3) The complete division of reference mechanism makes it easy to apply and improve.(4) Provided with varied forms of inference.(5) Also provided with the accurate condition of failure diagnose, as well as suggestions. By this model, the author adapted the Bayesian Network inference platform set by Decision Systems Laboratory, University of Pittsburgh, analysed some data, proved its efficiency and draw a outline design on inference diagram system as well.