Blade root bolt breakage failure detection method based on machine learning

Abstract

The invention provides a blade root bolt breakage failure detection method based on machine learning. The method comprises the steps as follows: known wind turbine data are preprocessed, and downsampling is performed on preprocessed wind turbine data; standard processing is performed on wind turbine data after downsampling, invalid features in the data after standard processing are deleted, and PCA dimensional reduction is performed on all the rest features; a multilayer perceptron model is constructed by the aid of data after dimensional reduction; and whether the current state of a wind turbine is the failure state is predicted by using the constructed multilayer perceptron model. The method has the beneficial effects as follows: with the method, the relatively high cost for manual detection in a wind turbine power generation plant can be avoided, and all that is required is to establish the model for the wind turbine data and detect whether failed/broken bolts exist by using the model.

Description

technical field [0001] The invention belongs to the technical field of artificial neural networks, and in particular relates to a machine learning-based fault detection method for blade root bolt breakage. Background technique [0002] In recent years, wind power generation equipment has increased significantly, especially wind power generators. However, in use, it is found that the blade root bolts often wear out too fast, and are prone to breakage due to industrial quenching problems, and these problems can easily lead to unstable power generation of wind turbines, and even failures such as blade shedding. During the operation of wind turbines, some bolt fractures must be dealt with on site, and even for regular maintenance, it is difficult to rely solely on manpower to find out whether the bolts have broken. The problem of bolt breakage has problems such as high maintenance cost and difficult maintenance. [0003] At this stage, the bolt failure problem of wind turbines...

AI Technical Summary

Problems solved by technology

However, these loading methods cannot transmit shear stress, or cannot simulate the friction behavior between bolts and connected parts in reality, and cannot consider the loss of preload caused by loose nuts.

As a result, in the actual finite element simulation process, the bolt stress generated is too large, so it is generally not used as a means of checking the bolt struct

Images