A Study On Life Assessment For High Temperature Components Based On Digital Image Processing

After a long term service of high temperature materials, creep damage occurs, which results in degradation of high temperature components. Therefore, accurate and reliable prediction of creep damage and remaining life provides a basis for timely, safe, and economic replacement or repair of key components, which helps to maximize a plant's usefulness by eliminating unnecessary replacements and reducing costly unscheduled outages caused by in-service failures.A-parameter method is a novel one based on creep cavitation to assess creep damage and predict remaining life. However, it is still implemented with people eyes under optical microscope in our country, which has many limitations such as human errors, time-consuming and energy-consuming. To overcome these limitations, based on replica inspections, combined with many technologies such as digital image processing, pattern recognition and artificial neural networks, A-parameter method is used in this paper to study creep damage and predicting residual life automatically. The main work is listed as follows:1. The state-of-the-art of life assessment techniques for high temperature component was introduced briefly. Then, the mechanism and methods of life assessment based on creep cavitation were reviewed in detail, and the digital image processing technology and its application were also discussed.2. The pre-processing for metallographic image is necessary to judge A-parameter accurately. The analysis for using image shadow remedy, enhancement, segmentation, the regions of interest extraction, grain boundary thinning and reconstructing, and image fusion to process the original metallographic images was carried out. The result shows the pre-processed images meet the requirements for judging A-parameter.3. The micro-cavities features are characterized by freeman technology as average gray level, roundness, and area. These features are employed to construct a BP neural network, which is a three-layer perception model and is learnt through BP and PSOalgorithm by samples to recognize cavitations automatically. The results show the classifier has higher accuracy for the clear image with coherent grain boundary.4. Based on ^-parameter method incorporated with image processing, pattern recognition and artificial neural network techniques, the life assessment software for high temperature materials with friendly user interface is developed under Visual C++.Net IDE.. A practical case to predict the HK40 reformer tube served for 69,000 hours, the residual life is pre-forecasted by the assessment system, and the results accord with the served station of material and meet the requirement.