Experiment And Prediction Study Of Water-cooled Radiator Pipe Erosion Wear

Water energy is the most widespread and the largest total energy,the use of hydroelectric generating units to develop and utilize water energy is of great significance.Hydro generator sets will produce a large amount of heat in operation,if not timely cooling,will damage the upper guide bearing,rotor and other components of the unit,and the radiator is a basic and important component of the power station cooling system,the role is to timely transfer the heat generated by the unit.As the water flow into the radiator pipeline with a certain amount of sediment,and for hydroelectric generating units,the flow rate of the radiator changes with the operating conditions,the long years of work so that the radiator pipe wall erosion wear phenomenon,which can seriously affect the normal operation of the unit,and even cause safety problems.Therefore,it is important to predict the erosion wear behavior of radiator pipes.This paper builds a water-cooled radiator pipe erosion wear test bench,selects purple copper bends as the research object,conducts tests on radiator pipes with different inlet speed,sand content and sand particle size,analyzes the sand-fluid-pipe inner wall collision mechanism,explores the development of wall mechanical properties and microscopic morphology under long time impact,applies numerical calculation methods to investigate the protection measures of erosion wear,and on the basis of the test machine learning method is introduced to establish PSOLSS VM model and PSO-BP model to predict the abrasion rate of radiator pipes.The main work and conclusions are as follows:(1)Due to the influence of the radiator pipe structure,the impact strength of the sand particles in the fluid on different parts of the inner wall of the pipe varies,As the angle of the outer arch side of the elbow increases,the severity of the erosion wear tends to increase first and then decrease.Under the action of inertial force,the sand particles collide with the pipe at a large impact angle,where the most serious part of the erosion wear behavior of the elbow is located at 60°on the outer arch side.The mass loss and abrasion rate of the bend specimen increased with the increase of inlet speed,the sand impact wall energy was small at low inlet speed,the abrasion rate increased slowly but was positively correlated with the test time,and the sand impact wear was mainly slight furrows and abrasion marks;the sand following enhanced at high inlet speed,the kinetic energy and impact frequency of sand-bend inner wall impact increased,and the material mass loss and abrasion rate increased significantly.The erosion wear is mainly deformation wear and flaking,and the content of sand particles embedded in the material surface and the depth of erosion pits will also increase significantly.(2)The change of sand content on the evolution of the radiator bend erosion wear has a catalytic effect,low sand content of the wall damage effect is close to the clear water conditions;with the increase of sand content,the number of sand particles impacting the wall increases thus intensifying the wear of the radiator pipe;when the sand content increases to a certain critical value,the surface roughness is a significant increase.The change of sand particle size on the material surface under the small particle size to the normal direction of the extrusion deformation,no obvious tangential force effect of cutting wear;large particle size under the abrasion rate has a drop type growth,the material surface to the tangential direction of the cutting effect is the main,accompanied by a certain amount of normal impact wear,at this time the etch pits connected into pieces.(3)Machine learning was used as the prediction analysis tool to establish four types of radiator pipe erosion prediction models:BP,PSO-BP,LSSVM and PSO-LSSVM prediction models,and the models were trained and tested using test data as the sample set.The mean absolute error,mean square error and mean absolute percentage error are much smaller than the other three prediction models,and the overall prediction of radiator pipe bend erosion test is more accurate and the network model is more effective.