Loose Parts Monitoring System Improvement

• Main Authors: Weng,Ming Cheng, 翁明誠
• Other Authors: Minsun Ouyang
• Format: Others
• Language: zh-TW
• Published: 1994
• Online Access: http://ndltd.ncl.edu.tw/handle/76770467191739267633

碩士 === 國立清華大學 === 動力機械學系 === 82 === Mass estimation has been an unsolved difficulty to Loose Parts Monitoring System since the lack of theoretical description and the huge scale of the reactor system which make a thorough understanding of the characteristics of impact signal of loose parts unreachable. Hence,analysis of the loose parts purely rely on expert's experience and the prediction is generally with large uncertainty .This paper apply multi-layer neural network on mass estimation.With the noise tolerance capability of neural network, mass estimation can be done without consult experienced expert and this technique can be adapted easily without changing the facility. Base on experimental study, the mass estimation of loose parts can be obtained with the variance of about 7%.