Brick Interior Defect Diagnosis Based On Sound

As a common industry material, refractory is used in many industries just as metallurgy, cement, glass, ceramics, energy power and some advanced branches of science, and plays an important role in development of national economy. Particularly for the speedily development of steel production in recent years, the used amount of refractory is more and more, its quality and service life influence safety production and economy benefit of steel industry directly.Nondestructive testing method based on sound has a long history; it's an effective and traditional quality testing measure, especially for interior quality testing. Striking method is widely applied as a result of simple operation and economical among so many kinds of means. In this paper striking method is adopted to test brick interior quality fault, according to brick striking sound it is judged if there are defects and differentiate fault types.It is inevitably that there are noises embedded in field collected signal, use infinite impulse response (IIR) digital filtering technique which computes rapidly and exactly to de-noising, and it meet real-time demand of online detection. On the other hand, striking sound is a series of interval signal, adopt approximate entropy computing judging start and ending points of striking sound, then get extracted signals, and this method is robust to high amplitude pulse noise.Extracted signals are time-domain signal without obvious regulation and cannot be used to detect defects. Apply Welch method estimate power spectrum and then detecting by spectrum data. However, there are so many data in power spectrum with large range, it is impossible using all the data. Nevertheless, peak points are energy concentrations in one signal, and there are correlations within peak points of the same spectrum, therefore propose Principal Component Analysis (PCA) detecting interior defect according to the correlation, show the power spectrum peak points extracting method as well. To the modeling bricks with known fault conditions, in the case of getting enough normal brick information, propose PCA to testing inner defects, build principal component model by peak points'data of normal brick and calculate statistic thresholds to detect; furthermore, to many types of interior defects, in the case of getting enough defective bricks information, put forward multi-PCA to diagnosis, multi-PCA models are built by defective modeling bricks to distinguish fault types. For modeling bricks with unknown information, fuzzy clustering is used to obtain defect information, take normal and each defect type brick as a cluster, unsupervised classify modeling bricks by fuzzy clustering, get defect types in the end.After these, take test simulation of these means, and gain better defect test and diagnose result.