| DC 欄位 |
值 |
語言 |
| DC.contributor | 工業管理研究所 | zh_TW |
| DC.creator | 蘇翊甄 | zh_TW |
| DC.creator | Yi-Jhen Su | en_US |
| dc.date.accessioned | 2020-7-2T07:39:07Z | |
| dc.date.available | 2020-7-2T07:39:07Z | |
| dc.date.issued | 2020 | |
| dc.identifier.uri | http://ir.lib.ncu.edu.tw:88/thesis/view_etd.asp?URN=107426028 | |
| dc.contributor.department | 工業管理研究所 | zh_TW |
| DC.description | 國立中央大學 | zh_TW |
| DC.description | National Central University | en_US |
| dc.description.abstract | 近年人工智慧的快速發展，促使了工業 4.0 的演進，現今大多數工廠都已漸漸導入智慧製造系統，也代表著機器的穩定度對工廠生產是相當的重要。以往工廠對設備的維護為修復性維護或預防性維護，此兩種方式對於維護成本是相當大的，且可能造成設備在無預警的狀況下發生異常停止，為了不讓設備在無預警的狀況下發生異常，近年來已漸漸發展為預測性維護。基於前述問題，本研究主要動機為期望在設備發生故障前能夠準確偵測到異常，發出警訊提醒設備人員以進行設備維護，目的為提早預防機台異常的發生而導致停止生產之狀況。
本研究所使用數據為 A 公司所提供之塗佈機感測器數據，分別使用長短期記憶網路 (Long short-term memory, LSTM) 模型以及門控循環單元 (Gated Recurrent Unit, GRU) 模型的監督式學習來建立機台異常偵測系統，並以多對多的滑動窗口方法來進行模型的訓練及預測，加速模型訓練速度及降低計算的複雜度，以及針對不同模型做超參數的調整，再利用評價指標方法找出最佳的模型配置。 | zh_TW |
| dc.description.abstract | In recent years, rapid development of artificial intelligence has promoted the evolution of Industry 4.0. Nowadays, most factories have gradually introduced smart manufacturing system, which also represents the stability with the machine is important for factory capacity. In the past, the maintenance of equipment in the factory was corrective maintenance or preventive maintenance. These two methods are quite expensive for maintenance and it may cause the equipment to shut down abnormally without any warning. In order to prevent the equipment to shut down without warning, predictive maintenance has gradually developed in recent years.
As stated above problems, the motivation is to expect to detected the anomaly before the machine shut down, and alert equipment engineer to maintenance of equipment. We will use the A company’s coating machine sensor data to analysis the anomaly detection.
We will use supervised learning method base on the Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) model to build anomaly detection system, and using slide window to speed up the training time and reduce the computational complexity. We experiment the various Hyperparameter for the different model, and using confusion matrix for performance evaluation the model to get the most suitable Hyperparameter to optimize the model. | en_US |
| DC.subject | 深度學習 | zh_TW |
| DC.subject | 時間序列 | zh_TW |
| DC.subject | 長短期記憶網路 | zh_TW |
| DC.subject | 門控循環單元 | zh_TW |
| DC.subject | 異常偵測 | zh_TW |
| DC.subject | 預測性維護 | zh_TW |
| DC.subject | Deep Learning | en_US |
| DC.subject | Time Series | en_US |
| DC.subject | Long Short-Term Memory | en_US |
| DC.subject | Gated Recurrent Unit | en_US |
| DC.subject | Anomaly Detection | en_US |
| DC.subject | Predictive Maintenance | en_US |
| DC.title | 基於 LSTM/GRU 於塗佈機之異常偵測 | zh_TW |
| dc.language.iso | zh-TW | zh-TW |
| DC.title | Anomaly Detection of the Coating Machine base on LSTM/GRU Approaches | en_US |
| DC.type | 博碩士論文 | zh_TW |
| DC.type | thesis | en_US |
| DC.publisher | National Central University | en_US |