| DC 欄位 |
值 |
語言 |
| DC.contributor | 資訊工程學系 | zh_TW |
| DC.creator | 曾翊銘 | zh_TW |
| DC.creator | Yi-Ming Zeng | en_US |
| dc.date.accessioned | 2018-6-12T07:39:07Z | |
| dc.date.available | 2018-6-12T07:39:07Z | |
| dc.date.issued | 2018 | |
| dc.identifier.uri | http://ir.lib.ncu.edu.tw:88/thesis/view_etd.asp?URN=105522116 | |
| dc.contributor.department | 資訊工程學系 | zh_TW |
| DC.description | 國立中央大學 | zh_TW |
| DC.description | National Central University | en_US |
| dc.description.abstract | 隨著相關技術的成熟，全球製造業結合物聯網、大數據、深度學習與雲端計算等先進技術，從工業3.0的「自動化」走向工業4.0的「智慧自動化」，藉由相關技術的導入，有助於提升生產的效率和穩定性，並提高工廠客製化的能力、以應對市場需求的變化。本論文探討工業4.0中的故障預測與健康管理（Prognostics and Health Management，PHM）研究，透過預估設備或零件剩餘可用壽命，在設備停止運作前，提早進行設備的維修或零件更換，以降低因設備損壞所造成的生產損失，提高生產線的可靠性。
本論文提出一個結合卷積神經網路與循環神經網路的深度學習模型，使用基於注意力機制的長短期記憶神經網路處理時序資料，可以透過LSTM記憶單元
儲存過去輸入的資訊，並且透過注意力機制，在不同的時間點，關注不同的特徵，進行更加準確的預測。本論文在模型前半部加入卷積層與池化層，用以解決原始資料時間序列過長所造成的問題。
本論文將所提方法應用於法國貝桑松研究機構FEMTO-ST的軸承資料集(PRONOSTIA)，以預估軸承的剩餘可用壽命，實驗結果顯示，無論是在均方差(Mean Squared Error, MSE)或均方根誤差(Root Mean Squared Error, RMSE)，本論文所提的方法都優於其他相關方法。 | zh_TW |
| dc.description.abstract | With the maturity of related technologies, global manufacturing combines advanced technologies such as Internet of Things, big data, deep learning, and cloud computing, from the ＂automation＂ of Industry 3.0 to the ＂wisdom automation＂ of Industry 4.0. through those technologies, it will increase the efficiency and stability of production, and increase the ability of factory customization to respond to changes in market demand. This paper focuses on the Prognostics and Health Management (PHM) study in Industry 4.0. By estimating the remaining usable life of equipment or parts, equipment repairs or replacements can be performed before it stopped.
This paper proposes a deep learning model that combines convolutional neural networks and recurrent neural networks. It uses long short term memory neural networks based on attention mechanisms to process temporal data, it can store past information through the LSTM memory unit, and through The attention mechanism, our model can focus on different feature at different time and makes more accurate predictions. convolution layer and pooling layer are added in the first half of the model to solve the problems caused by the long sequence.
In this paper, the proposed method is applied to the bearing data set (PRONOSTIA) of FEMTO-ST, a research institute of Besançon, France, to estimate the remaining useful life of the bearing. The experiment results show that both mean squared error (MSE) and Root Mean Squared Error (RMSE), the method proposed in this paper is superior to other related methods. | en_US |
| DC.subject | 智慧工廠 | zh_TW |
| DC.subject | 剩餘可用壽命 | zh_TW |
| DC.subject | 深度神經網路 | zh_TW |
| DC.subject | 長短期記憶 | zh_TW |
| DC.subject | 卷積神經網路 | zh_TW |
| DC.subject | 注意力機制 | zh_TW |
| DC.title | 卷積注意力機制長短期記憶深度學習 用於軸承剩餘可用壽命預估 | zh_TW |
| dc.language.iso | zh-TW | zh-TW |
| DC.title | Convolutional Attention-based Long Short-Term Memory Deep Learning for Estimating Bearing Remaining Useful Life | en_US |
| DC.type | 博碩士論文 | zh_TW |
| DC.type | thesis | en_US |
| DC.publisher | National Central University | en_US |