Abstract
核能電廠鬆動元件監測系統估計質量之問題因為理論對此缺乏完整的描述因此只能依賴專家之經驗判斷鬆動元件事件之嚴重程度, 而一般估計的結果其準度不確定且時效性較低。本文主要是探討如何應用類神經網路於鬆動元件監測系統之質量估計以及判斷位置時需注意的事項, 使用類神經網路可避開理論和量測上之困擾, 由本文之實驗可驗證，利用類神經網路技術, 結果是在不需更動, 增加硬體設備且無須諮訊專家的情況下，估計的標準差可達7%。Mass estimation has been an unsolved difficulty to Loose PartsMonitoring System since the lack of theoretical descriptionand the huge scale of the reactor system which make a thoroughunderstanding of the characteristics of impact signal of looseparts unreachable. Hence,analysis of the loose parts purelyrely on expert's experience and the prediction is generallywith large uncertainty .This paper apply multi-layer neuralnetwork on mass estimation.With the noise tolerance capabilityof neural network, mass estimation can be done without consultexperienced expert and this technique can be adapted easilywithout changing the facility. Base on experimental study, themass estimation of loose parts can be obtained with thevariance of about 7%.