The Methods For Detection And Classification Of Steel Ingot Surface Defects

In recent years, the continuous and rapid growth of china's economy lead to the sharp increase of the domestic steel consumption. Ingot is one of the various kinds of steel. Most of the steel used today is made of ingot. Nowadays, in the years of the rapid growth steel industry, the pursuit of producing high quality ingot is the major goal for the giant steel enterprises. The quality inspection is one of the essential sectors in the ingot production. Only when the producer inspects their product accurately and timely, can they adjust their production strategy according to their specific production situations. The effective inspection of the ingot can guarantee the effectiveness and the quality of the production as well as the enterprise's economic benefit.The molten steel is poured inte molding solidification to form ingot. Ingot casting is divided into two methods:top casting and bottom pouring. Top casting, in general, ingot internal structure is better, less inclusion, and the operation fee is low. Bottom pouring, ingot surface is good in quality, but molten steel pour in feed trumpet make steel inclusions increase.Casting temperature should be strictly controlled. If casting temperature is too low, when the steel liquid was poured into the mold ingot surface defects, when liquid steel is in the barrel, it begin to solidate, which will cause metal loss or the whole furnace steel scrap; If casting temperature is high, it will delay the formation time and the surface of ingot has hot cracking.Common defects are:ingot surface scar, heavy leather and the longitudinal and lateral crack, the internal residual shrinkage, subcutaneous air bubbles, loose and segregation, in-clusion, etc. These defects can greatly reduce the ingot into billet rates, even make whole ingot scrapped. We mainly will these defects are divided into four types:corner cracks, porosity, central segregation, central loose.This paper mainly apply digital image processing techniques to analyse and detect the image of the defect in ingot instead of traditional artificial detection methods. According to different defects types it put forward different testing methods. First is the pretreatment of the images, which includes:Gray standardization, image denoising (median filtering, gaussian low-pass filter) and binary, etc Through these processing, the image quality can be improved, and The images of defect feature information can appear.Next we draw the feature extraction of ingot, first we calculate the image texture statis-tics, we divided the image into 35 x 35 chunks and calculate average gray, smoothness, consistency, extract the features of shrinkage defect of shrinkage, will again break. Af-ter denoising small area The affecting judgment, get rid of the small area defects. This is mainly divided into two steps:calculation area to remove a small area noise and morpholog-ical processing, morphological processing including expand, corrosion, etc. Through this processing make crack defects is more obvious. Reuse hough transform, and the red line will line detection crack location marked. Among them, in this part of our solved. kirch-hoff transform given a comprehensive theory knowledge, as well as the realization of the algorithm and experimental results, thus will defect feature extraction out of crack.In the end, we analyze the feature of the center segregation and center porosity to make an comparison. Firstly, we calculated the binary image, in which 8 connection to the number of connected domain, calculate these connected domain Average area. Through calculation and a lot of experiments And we did get center segregation and center two kind of defects of osteoporosis in the scope of the average area area The defects, and then will they tell apart.This paper categorizes the four defects by the five steps above and transfer their features of the four defects into math language which can be identified by the computer. This paper applies MATLAB to conduct large amount of experiments and data analysis, which draws a satisfying result. This detect method can meet the demand of the detection of most of the ingot production lines and it is also easy to apply so the application value is high.