Condition Recognition And Quantitative Analysis Of Internal Leaks Through Valves Based On Infrared Thermography Method

Valve plays an important role in oil and gas industry which can control fluid in pipeline or storage tank.Once valve has leakage,it not only affect the normal operation of technological process,but brings environment pollution and casualty.Therefore it is necessary to take effective measures to control the situation so as to ensure safety.Infrared thermography based on surface temperature difference of aim object to do condition identification visually and quickly.The valve surface temperature will redistribution after leakage which reflects in infrared thermal image used to judge whether the valve leak or not.So there is important significance to carry out the valve leak detection technology research.After valve internal leakage three-dimensional model establishment,flow field and temperature field under different process parameter,medium are simulated.Diverse factors such as internal leakage clearance,pressure,temperature,medium are simulated which can provide theoretical guidance for experimental design.With the help of fluid leakage and diffusion test system(China university of petroleum(east China)),Infrared detection experiment device is setted up.Then experiments study on the effect of leak rate,pressure,temperature are carried out by valve test equipment.Valve state judgement is performed by analyzing the infrared thermal images.Meantime,Temperature characteristic value is extracted from images to do parameters impact analysis.Temperature feature signal of different failure mode are extracted.Then BP neural network and support vector machine are used to the identification of failure mode respectively.Meanwhile,the performance of grid search,particle swarm optimization,genetic algorithm are analysed.The results show that support vector machine can meet precision requirement with grid search parameter optimization method and the mode accuracy,training accuracy and testing accuracy are 97.019%?99.46% and 99.18%.So it can be used to pattern prediction for unknown valve leakage.The leakage rate is calculated by heat transfer model under various experimental factors.Then BP neural network and support vector machine are used to the regression analysis.The results presents that BP neural network has better performance.Formula fitting of leakage rate is established depending on least square method after Single factor effect analysis.The risk matrix is builted with the leakage impact index as x-coordinate and leak probability as y-coordinate.At last,the valve internal leakage detection system sofrware based on infrared thermography is designed to guide safety management for engineering field.