Font Size: a A A

Research On Measurement And Simulation Of Neural Network Model For Dynamic Characteristic Dimension Of Spray Columnar Deposit

Posted on:2009-10-27Degree:MasterType:Thesis
Country:ChinaCandidate:Z T MaFull Text:PDF
GTID:2121360248952131Subject:Materials Processing Engineering
Abstract/Summary:PDF Full Text Request
The final shape of deposits is an important parameter of spray forming process,there is great significance to control the shape during the manufacture process.In order to improve the precision of deposit dimensions,investigation on the detecting and modeling of deposit dimensions in spray forming are carried out in this study.Base on the existence spray forming equipment,the designed hardware system can transfer the deposit dimensions expressed by optical information difference in digital form into the computer.At the same time an excitation motor PLC real time timing system has been build,which can shift the sediment decline speed in different experimentation process. A simple camera calibration technique is introduced to establish the conversation relationship between the images coordinates and the work piece coordinates.Base on VC++,an image processing algorithm is developed to calculate the deposit dimensions:an pre-treatment algorithm is used to improve the image quality further;two pole Zemike moments operator is used to fast detect the deposit edge accurately.Base on the Bezier curve,repairing the previous rude edge which can eliminate broken points and make the edge more smoothness.The testing result shows validity of the image measurement algorithm:the running time is too short to disturb the detecting time,and the average relative error is less than 0.97%,meeting the requirements on the real time ability and high precision of the detection of deposit dimensions.A three-layer BP neural network model,which consists of 12 cells in the input layer,60 cells in the hiding layer and 4 cells,in the output layer,is established as the deposit dimension model.The testing shows the neural network model can describe the spray forming process with high accuracy:the average output error is 4.08%and RMS is 3.3398x10-3 mm.The simulation of the neural network model is done to obtain the variation rules of the characteristic deposit dimensions.The height and margin between two diameters of spray forming deposit increase all the time during the spraying process; increasing the pulverization gas pressure in the propriety scope,the height and margin between two diameters of spray forming deposit increase at first and then decrease,the height reach the top when pressure equal to 0.9MPa and speed equal to 46mm/min,the margin between two diameters touch bottom when pressure equal to 1.OMPa;increasing the descend speed in the same pressure,the height and margin between two diameters increase,the situation differs in different pressure.
Keywords/Search Tags:Spray forming, Deposit dimension, Image processing, Neural network
PDF Full Text Request
Related items