A fault identification method for induced draft fan based on cnn-svdd

Abstract

The invention relates to a CNN-SVDD-based induced draft fan fault identification method, comprising: step 1: collecting sufficient training data, and performing data preprocessing; step 2: constructing a CNN-SVDD model using the pre-processed training data, and using The algorithm reduces the dimensionality of the time series data, and then uses the SVDD algorithm to classify the data; Step 3: Collect sufficient real-time data, and use the constructed CNN-SVDD model to identify the fault of the induced draft fan corresponding to the real-time data. The beneficial effects of the present invention are: the present invention uses deep learning to perform feature extraction first, and then uses SVDD to perform single-category analysis after greatly reducing feature dimensions, which can save time and memory overhead of SVDD. In addition, due to the time series characteristics of induced draft fan data, the local correlation in the time dimension of time series data can be fully utilized, and the one-dimensional convolutional network is used to reduce learning parameters and further reduce computing overhead.

Description

technical field [0001] The invention relates to the technical field of reliability maintenance engineering, and particularly includes a CNN-SVDD-based induced draft fan fault identification method. Background technique [0002] The induced draft fan is an important auxiliary equipment of the thermal power generation unit. The operating characteristics of the induced draft fan will affect the boiler combustion efficiency of the unit, the operation stability of the unit and the output of the unit. Therefore, timely detection of various fault symptoms during the operation of the induced draft fan can effectively grasp the operating status of the induced draft fan and optimize its operating status, avoid further expansion of the fault trend and the occurrence of unplanned shutdown events, and improve the overall performance and efficiency of the unit. , It is of great significance to reduce the maintenance cost of equipment and increase the available time of equipment. [0003]...

AI Technical Summary

Problems solved by technology

In the past, fault diagnosis methods based on machine learning modeling were mostly aimed at a single power plant, using all historical fault data of specific equipment in the power plant for fault pattern recognition model training, but fault pattern recognition modeling based on machine learning requires a large number of fault sample data to Ensure high diagnostic accuracy

In fact, the sample data of the same fault case of similar equipment in a single power plant is limited, and large diagnostic errors will occur in the actual application process

Images