Fault Diagnosis Method of Engine Turbine Disc

Abstract

The invention provides a fault diagnosis method for an engine turbine disk, comprising: receiving a start instruction, the start instruction is used to instruct to initialize a training data set, initializing the training data set based on the start instruction; calculating the weight of the samples and training the weighted support vector through the training data set The machine model obtains the support vector data set, and stores the support vector data set; collects new test data blocks, uses the weighted support vector machine model to perform fault detection on the sample, and detects whether the sample is a fault sample; if the detected sample is a fault sample, a fault occurs Alarm signal; after receiving the end detection command, the fault detection is ended based on the end detection command. The invention based on the incremental learning-based time series fault detection method of the engine turbine disk and the weighted support vector machine adaptive update method based on the incremental and decrement modes effectively solves the drift problem of the time series data flow and improves the reliability of the fault detection model. The calculation speed keeps the correctness of fault detection results.

Description

technical field [0001] The invention belongs to the field of civil aircraft aero-engine turbine disks, in particular to an engine turbine disk fault diagnosis method based on an adaptive weighted support vector machine. Background technique [0002] The operation process of the engine turbine disk is a non-stationary process including conceptual drift and unbalanced distribution. In order to monitor the operation process of the engine turbine disk through a data-driven model, it is necessary to study the non-stationary data flow, especially the key characteristics of the data flow such as conceptual drift and unbalanced distribution. [0003] In the prior art, the literature Lee Y H, Jin H D, Han C. "On-Line ProcessStateClassification for Adaptive Monitoring", Industrial&EngineeringChemistryResearch, 2006,45(9):3095-3107 introduced a process monitoring adaptive based on knowledge rules This technology uses if-then rules to extract process knowledge and realize the detection...

AI Technical Summary

Problems solved by technology

[0003] In the prior art, a knowledge rule-based process monitoring adaptive Technology, which uses if-then rules to extract process knowledge and realize the detection of operating status. The experimental results have verified that the adaptive monitoring method can significantly reduce the false alarm rate (FAR), but the fuzzy processing of information will lead to the reduction of system control accuracy.

[0004] According to the literature He X B, Yang Y P. "Variable MWPCA for Adaptive Process Monitoring", Industrial & Engineering Chemistry Research, 2008, 47(2): 419-427. He and Yang proposed a time-varying process monitoring method based on moving window PCA. Combining window technology and R-SVD algorithm, an adaptive fault detection model is established. This method can effectively improve the performance of the adaptive fault detection method, but redundant data can easily affect the analysis results.

Images