Design And Implementation Of Online Hardness Detection System Based On Barkhausen Noise

This paper presents a design scheme of hardness on-line measurement system based on Barkhausen noise.In order to meet the market requirements for improving the efficiency,convenience and accuracy of hardness measurement,a new hardness detection method based on frequency domain features is used.Combining machine learning and Internet technology,a complete set of system including on-line hardness detection and data management functions is realized.The innovation of this paper is mainly embodied in the design of hardness detection algorithm,including a set of new features better than traditional ones,a set of new feature processing methods,and the use of GBDT model to complete the learning process.The main work includes the following parts.Firstly,the requirement of the system is analyzed.According to the result of requirement analysis,the structure and function of the system are determined.Finally,the technical scheme of the system is studied and analyzed,and the technical selection is determined.Then the detection method based on frequency domain features is studied,and the problems and limitations of time domain features are analyzed.To solve these problems,the scheme of using AR spectrum of Barkhausen noise signal as feature source is proposed.The AR spectrum of Barkhausen noise signal is studied and analyzed,and a new set of AR spectrum characteristics based on AR spectrum are proposed,including the peak value in AR spectrum of Barkhausen signal,the peak valley width in firstorder difference of AR spectrum of Barkhausen signal,the valley point value in secondorder difference of AR spectrum of Barkhausen signal and the location of valley in second-order difference of AR spectrum of Barkhausen signal.Based on the extracted features,a set of new feature processing methods based on K-means algorithm and Relief algorithm are proposed.Firstly,k-means algorithm and One-Hot coding are used to unify and express the feature dimension,and then Relief algorithm is used to select and process the sample features.After feature processing,GBDT model is used as a tool to learn the relationship between feature and hardness,thus the design of hardness detection algorithm is completed.Finally,2700 samples collected from Cr12 MoV material are used as data sets to verify the practicability of the proposed method.Experiments show that the error value of the frequency domain feature detection method is 6.3,which is less than 14.2 of the time domain feature detection method and 9.8 of the traditional frequency domain feature detection method.This shows that the method is better,thus proving the effectiveness and superiority of the new frequency domain feature detection method in the data of this subject.Finally,the online hardness detection system based on Barkhausen noise is developed,using Django as the system development framework and Mysq as the database.The specific development work includes the design of database model and data table,the construction of background server,the compilation of algorithm module and business module,and the development of front-end page of the system website.Tests show that the system has the advantages of low cost,easy storage and query of data,convenient and accurate hardness testing,and meets the needs of hardness testing.The system runs stably and has high practical application value.