| Large forgings are key parts of major equipments which have been widely used in electricity, aerospace, ships, heavy machinery and other fields, its quality directly affects the overall level and operation reliability of equipments, it is necessary to detect its possible internal flaws such as cracks, porosities and inclusions. Ultrasonic testing has many advantages such as strong penetrability, high flaw detection accuracy, high sensitivity, low detection cost and harmless to human beings and environment, so it is suitable for flaws detection of large forgings. At present the ultrasonic testing of large forgings are mostly done by manual scanning and human readable interpretation in domestic, it easily leads to missing detection and misjudgment, so detection efficiency is low and reliability is poor, it is necessary to develop an automatic ultrasonic testing system of large forgings, thereinto testing method and signal processing are key technologies the system involved, but current technologies have some limitations, in this dissertation, these technologies were investigated theoretically and experimentally, the main research contents were as follows:1. A multi-channel automatic ultrasonic testing method suitable for large cylinder forgings and large medium plates was put forward. For large medium plate main ultrasound probes (double as transverse edge probes) were placed in the middle of transverse edge, longitudinal edge probes were placed in two longitudinal edges, when detecting transverse edges, plate was fixed and main probes swang side-to-side, when detecting other parts, plate made uniform linear motion along the rolling direction, main probes swang side-to-side and longitudinal edge probes were fixed, so the scanning track of plate edges was rectangle and plate body was sine or cosine; For large cylinder forgings straight beam probes and angle beam probes were respectively longitudinally placed in two lateral wall of it in diameter direction, when detecting, cylinder forgings revolved around its axis in situ and at the same time two groups of probes moved along the axis, so the scanning track was spatial spiral path. Ultrasonic transducers grouped and placed in different positions automatically scanned workpiece simultaneously in all directions in specified path which improved detection efficiency and reliability.2. The de-noising theories and algorithms of ultrasonic echo signal were investigated. An ultrasonic signal de-noising algorithm based on wavelet transform and independent component analysis was proposed (WICAW). The original signal was decomposed by wavelet transform, ICA algorithm was applied to decomposed coefficients and isolated independent components were assessed by threshold value to filter out noise, de-noised ultrasonic signal was obtained by reconstructing wavelet coefficients. Simulation and experimental results showed that the proposed algorithm didn’t lose useful information and improved signal-to-noise ratio, performance was superior to wavelet soft threshold de-noising algorithm.3. The feacture extraction and recognition technology of defect ultrasonic signal was investigated, a feature extraction and identification algorithm of flaw ultrasonic signal based on wavelet coefficients clustering and support vector machine was proposed. Firstly, de-noised ultrasonic echo signal was decomposed by wavelet transform, decomposed wavelet coefficients were clustered by the method of probability statistics clustering, and finally the wavelet coefficients energy of each clustering was calculated and input to support vector machine classifier to identify flaws. The experimental results of blocks with typical flaws showed that the proposed algorithm reduced the calculation time of classifier and improved flaw recognition accuracy with small samples.4. An automatic ultrasonic detecting system of large forgings based on Ethernet was designed, the experiment platform was constructed and experiments were carried out. The validity of proposed testing method and signal processing algorithms was verified. |
