A Weighted kNN Fault Detection Based on Multistep Index and Dynamic Neighborhood Scale Under Complex Working Conditions

Abstract

Fault detection based on k-nearest neighbor (FD-kNN) is one of the most widespread fault detection techniques for industrial processes under complex working conditions, owing to its characteristic of local modeling. However, its state separation ability tends to worsen when the operating data is heterogeneous distribution. To tackle this challenge, a weighted k-nearest neighbor fault detection method based on multistep index and dynamic neighbor scale is proposed. The multistep nearest neighbor index is defined to evaluate the state separation ability, and a weighted k-nearest neighbor fault detection framework is formed by the assigned weights obtained from kernel principal component analysis. On the basis above, a dynamic neighborhood scale correction method and a dynamic threshold setting strategy are proposed to deal with the heterogeneous distribution of operating data and track the abrupt change of the operation state. 10 common faults of wind turbines with complex operation conditions are used to verify the effectiveness of the proposed method.

Wind turbine condition monitoring based on double‐layer ensemble KNN method

“…Instead, they directly utilize stored data to address classification or regression problems, demonstrating the potential for flexible handling of massive amounts of data. 8 Nonparametric modeling methods include multivariate state estimation technique, 9,10 Auto-Associative Kernel Regression, 11 and K-nearest neighbor (KNN). 12 KNN is a classical nonparametric method often used to solve classification and regression problems.…”

“…Nonparametric methods do not require the establishment of general inference formulas based on samples. Instead, they directly utilize stored data to address classification or regression problems, demonstrating the potential for flexible handling of massive amounts of data 8 …”