Application of maximum entropy probability density estimation approach to constituting oil monitoring diagnostic criterions

New diagnostic criterions based on maximum entropy principle for assessing wear condition are presented. By the new measure of critical boundary points defined in this paper, the precision of assessing wear condition can be promoted. Applying maximum entropy principle to get the most unbiased probability distribution of monitoring data can improve the accuracy of estimating probability distribution and no assumption of prior distribution is needed. These criterions are applied in our monitoring projects and their performances, especially in Spectrometric analysis and Direct Reading Ferrographic analysis field, are very good. According to the discussion of two examples, the method of constituting criterions based on maximum entropy probability density estimation approach is expatiated and the practicality is validated. (c) 2005 Elsevier Ltd. All rights reserved.