Fan fault transferable diagnosis method based on data enhancement and capsule neural network

Abstract

The invention discloses a fan fault transferable diagnosis method based on data enhancement and a capsule neural network. The method comprises the following steps: preprocessing collected fan vibration signal data, and detecting and eliminating abnormal values; extracting an optimal characteristic frequency band of the fault based on the average power spectral density; calculating the average power spectral density intensity value of the fan vibration signal on the optimal fault characteristic frequency band, and taking the average power spectral density intensity value as the input of a first-class support vector machine to carry out fault degradation detection so as to determine the initial failure occurrence point of the fault; the vibration signals are re-divided into fault data and normal data according to failure points, the data are labeled, and a training data set is constructed; initializing network hyper-parameters of the capsule neural network, and training the network hyper-parameters; and inputting a new vibration data signal into the trained network to obtain a diagnosis result. According to the method, the fault samples are effectively expanded through data enhancement, and the accuracy and mobility of model fan fault diagnosis are improved based on the multi-dimensional rich features extracted by the capsule neural network.

Description

technical field [0001] The invention relates to the technical field of wind power generation and the technical field of fault diagnosis, in particular to a transferable fault diagnosis method for wind turbines based on data enhancement and capsule neural network. Background technique [0002] In recent years, countries around the world have been responding to climate change and the near-depletion of fossil energy by supporting and investing in renewable energy. Among them, wind power generation has received great attention because of its natural, renewable and non-polluting characteristics. As the number of wind turbines continues to grow, wind turbines are becoming more expensive to operate and maintain, and attention has shifted to developing turbine condition monitoring and preventive maintenance technologies. [0003] Existing wind turbine fault diagnosis methods can be roughly divided into two categories: traditional machine learning methods based on mathematical models...

AI Technical Summary

Problems solved by technology

Although the second method is less demanding on human expertise, it requires a large amount of industrial field data as a support

In addition, because wind turbines are often installed in locations with harsh weather and steep terrain, wind energy companies cannot often send relevant personnel to inspect the operation status of wind turbines. They can only conduct regular inspections and record the operation status of key components. date, rather than the date the failure started, resulting in limited failure data

Images