Hardness Prediction Algorithm of Ferromagnetic Material Based on Improved Algorithm of BP Neural Network

Abstract

The hardness prediction algorithm of ferromagnetic materials based on the improved algorithm of BP neural network first collects the Barkhausen signal of ferromagnetic materials, divides the signal set, and obtains the Barkhausen noise training set and Barkhausen noise test set. Then analyze the AR spectrum of the collected signal, select 5 orders to expand, which are 4, 8, 16, 32, and 64 orders, and calculate the second-order derivative of the expanded signal, and use the valley width of the second-order derivative signal, the valley width The position of the deep and valley points is used as a feature, and the kmeans algorithm is used to carry out the distance for these valleys, and the signal is encoded, thus completing the unification of the feature dimensions. Then optimize and train the BP neural network model. The simulation shows that the prediction result of the present invention is very good, the mean square error is only 80, that is, the error of each hardness prediction can be guaranteed to be 9 Vickers hardness, and the mean square error of the time domain algorithm is 229, that is, it is greater than 15 dimensions Its hardness proves the effectiveness of the algorithm.

Description

technical field [0001] The invention relates to a non-destructive detection algorithm for ferromagnetic materials. According to Barkhausen noise, the traditional time-domain characteristics and BP neural network are improved, and a ferromagnetic material hardness prediction algorithm with higher accuracy is designed, which belongs to regression analysis and non-destructive detection. Related areas. Background technique [0002] In the machinery, automobile, aerospace, petrochemical, national defense, military and power industries, the monitoring and prediction of component fatigue life is crucial. The microstructure of the material is one of the important factors affecting its operating life . Therefore, rationally controlling the production, processing and use of materials and reducing structural defects are important measures to ensure and prolong the service life. The hardness of ferromagnetic materials depends on its structure, that is, the change of the internal micro...

AI Technical Summary

Problems solved by technology

[0006] Third, it has a full process. The destructive test can only test the raw materials. After the finished product is made, whether it is before leaving the factory or in use, unless they are not prepared to continue to serve, destructive tests cannot be carried out, and the steel cables of the bridge , rails and other facilities, people just want to test whether they can continue to serve. After the damage detection is destroyed, they will be scrapped regardless of whether they can continue to serve. Therefore, non-destructive testing that does not destroy the performance of the test object is the only method that can be used.

[0008] The shortcomings of the existing methods: on the one hand, the traditional time-domain characteristics will be affected by other properties of the metal (such as temperature, residual stress, plastic deformation, etc.), so that the time-domain characteristics cannot form a single variable problem with the hardness, and finally On the other hand, because the initial weight of the BP neural network will affect the final training result of the neural network, and the initial weight of the BP neural network is randomly set, so the final training result of the BP neural network is not Stablize

Images