A fault detection method based on orthogonal local preserving projection

Abstract

The invention discloses a fault detection method based on orthogonal local preservation projection, belonging to the field of industrial monitoring and fault diagnosis. The method comprises the following steps: collecting process data under normal working conditions of an industrial system to form a training set; adopting Maximum Likelihood Estimation Method to determine the eigendimension. adopting The orthogonal local preservation projection method to compute the projection matrix and construct the monitoring statistics. adopting The kernel density estimation method to determine the threshold. acquiring system real-time process data as a test sample, calculating the statistics of the sample, and compared with the threshold, to determine whether there is a fault. The method of the invention is based on data modeling, does not need accurate analytical model, can keep topological structure and local information between data, has simple algorithm and is easy to realize, and can be widelyapplied in the fields of high-speed railway, chemical industry, manufacturing and the like.

Description

technical field [0001] The invention belongs to the field of industrial monitoring and fault diagnosis, and in particular relates to a fault detection method based on orthogonal local preservation projection. Background technique [0002] With the improvement of people's requirements for the safety and reliability of industrial systems, the problems of process monitoring and fault diagnosis have become the focus of scholars' attention. Traditional data-driven methods mostly use multivariate statistical analysis techniques, such as principal component analysis, partial least squares, independent component analysis, etc., and are widely used in industrial systems. However, these methods aim at global optimization, and cannot maintain the topological structure between data, and lose the local structure information of data, resulting in unsatisfactory fault detection results. The idea of ​​manifold learning can solve this defect, while the data is reduced from a high-dimensiona...

AI Technical Summary

Problems solved by technology

However, the projection vector of the LPP algorithm is not orthogonal, which makes it difficult to reconstruct the data

In addition, the eigendimension is an important parameter of the LPP algorithm, which seriously affects the accuracy and reliability of the algorithm

The eigendimension estimation is too low, and important data features will be projected into the same space; the eigendimension estimation process will cause the projection vector to be unstable

Images