Detection of Induced Damage in Medium-Density Fiberboard Panels Using a Neural Network Method

本研究針對類神經網路使用來偵測中密度纖維板結構低破壞程度之可行性評估。類神經網路是一種計算系統，它使用大量簡單的相連人工神經元來模仿生物神經網路的學習能力，其中又以由三層網路所連結之倒傳遞類神經網路(back-propagation neural network)最為被普遍應用在診斷、預測功能上。先前研究已成功的辨識出中密度纖維板在其彈性限度內受不同載重後，因其結構受力後所導致應力波頻譜之變化，延續此研究成果利用此頻譜來訓練倒傳遞類神網路，以非破壞應力波量測試中密度纖維板當受不同人為加工後（如表面開槽、中間層開孔等）的結構破壞程度之頻譜，經過學習之類神經網路，可以成功地偵測。

關鍵字

類神經網路；倒傳遞類神經網路；中密度纖維板；應力波

並列摘要

This research assessed the feasibility of using a neural network to detect induced and interior damage to small samples of medium-density fiberboard (MDF). The neural network was a 3-layer back-propagation network. The undamaged stress wave frequency spectrum patterns were used to train the neural network. In a previous study, we successfully used the trained patterns to evaluate low levels of damage in samples of MDF onto which various percentages of their estimated failure loads were applied. In this experiment, after introduction of grooves on the surface or a hole through the center of the samples, a small change in the wave patterns occurred. The neural network has the unique ability to train itself using data to recognize spectral patterns and was successfully used to detect structural damage.

並列關鍵字

neural network ； back-propagation neural network ； medium-density fiberboard ； stress waves

國際替代計量

Detection of Induced Damage in Medium-Density Fiberboard Panels Using a Neural Network Method

全文下載

主題瀏覽