Study On Fatigue Crack Detection Of Cylinder Head Materials Based On Image Processing

As a green industry,remanufacturing engineering has opened a new way for China's resource reuse and environmental pollution control,and has great potentials for development.Remanufacturing is a typical discrete manufacturing industry with high labor cost,high labor intensity and low degree of automation.As an important branch of manufacturing,intellectualization is an important future development direction.Engine cylinder head is one of the important parts of engine and belongs to typical hi-strength and thin-section complex casting,whose craft is complex and manufacturing cost is high.It has a very high value in automotive engine remanufacturing,which is one of typical parts of engine remanufacturing.As for the damage of cylinder head,80 % of the damage is concentrated in bridge zone of cylinder head.At present,there are two main methods used for the detection of the cylinder head blanks: magnetic particle inspection and bulge test.Both of these methods require manual operation,low efficiency,cumbersome process,and can't be tested online at present,which is inconsistent with the fast and efficient concept of modern enterprises.In this paper,a method based on image processing is proposed to detect crack in cylinder head.This method has advantage of non-contact detection,image processing equipment simple,easy to operate,low cost,saving labor,accuracy and reliable.Furthermore,the advantage of on-line detection establishes a good foundation for the realization of cylinder head manufacturing automation and intelligent.Crack detection for cylinder head based on image processing is mainly by means of machine vision system and MATLAB software.The machine vision system acquires high quality of the original image of crack.MATLAB software is implementation of the original image preprocessing,image segmentation and feature extraction.In view of the corresponding image processing algorithm,the experiment software system was designed,the crack on the surface of cylinder head material static image processing and feature extraction was completed,and the evaluation system of surface crack defects detection effectiveness was given.The research results were as follow:(1)Image gray distribution based on image gray scale histogram was analysed,and distributed in the area of the crack and the background.The gray value of the crack is small and the background gray value is large.The crack is mainly concentrated in the range of20~75 gray values,while the background is mainly within the range of 75~225 and is in the majority.The gray value contrast of the crack area is not obvious.The gray value of the background and the crack is partially overlapped.(2)Based on energy analysis of crack image and using Fourier spectrum method for noise interference analysis,the center is high frequency and the surrounding is low frequency.Spectrum diagram can be seen that crack area exists obvious grainy black and white point,random,which belongs to the typical salt and pepper noise,as for the whole image,it should be composition of the salt and pepper noise and gaussian noise,with salt and pepper noise majority.The RGB channel of the image was analyzed.And the channel of R and G are mainly composed of salt and pepper noise,while the channel of B was mainly composed of gaussian noise.The whole image is the superposition of two kinds of noise.(3)The pretreatment of noise enhancement for crack image is made.Two kinds of smoothing filters,neighborhood filter and median filter,are used to process and contrast the salt and pepper noise in the crack image.The denoising effectiveness of the median filter is proved by the nonlinear signal in the real world.It also points out that the median filter will smooth the image while de-noising,so that the fuzzy quality of the image will be reduced and the details will be lost.(4)A modified median filtering algorithm was proposed: adaptive median filter in this paper.The rectangular window Sxy definition of m×n image based on the statistical properties of the filter area,can handle a greater probability of impulse noise,such as salt and pepper noise,which can retain the details in smoothing the impulse noise.(5)The crack skeleton can be obtained by means of pixel conversion to obtain the crack length.Compared with actual measurement 17.52 mm,error value is merely 0.0188.The measurement data is accurate.