基於數據驅動之冰機效率及運轉參數建模與設備故障偵測

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Electrical Engineering > Electronic Thesis & Dissertation > Item 987654321/93325

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/93325

Title:	基於數據驅動之冰機效率及運轉參數建模與設備故障偵測
Authors:	謝育祥;Hsieh, Yu-Xiang
Contributors:	電機工程學系
Keywords:	非線性迴歸模型;故障分析;數據驅動;專家規則;non-linear regression model;fault analysis;data-driven;expert rule
Date:	2023-07-24
Issue Date:	2024-09-19 16:53:51 (UTC+8)
Publisher:	國立中央大學
Abstract:	本論文以冰水主機設備（冰機）節能維護為研究主軸，冰機在暖通空調系統中是非常耗電的設備，此種大型設備的效率提升與參數優化，往往需藉專業人員的經驗以及知識才能達成。本研究嘗試藉由冰機的運轉數據資料來分析與調整適合的冰機特徵參數，若無法調整則找出冰機設備故障因素，期許本研究對於冰機設備的維護貢獻一份心力。本研究以訓練資料訓練非線性迴歸模型(Non-linear regression model)，模型的參數是感測器收集到的參數資料。在模型設計上不直接結合冰機的物理性質做開發，發展數據驅動(Data driven)模式的研究。由於並不是所有參數都是可控的，可控參數被設定為其中四個，而系統在冰機效率低落診斷上給出的建議調整參數不限於可控參數，會包含到其他非可控參數，因此本研究針對此問題做進一步探討。在系統給出建議調整參數為非可控參數時，則透過訓練資料搭配最小平方法(Least square algorithm)來建立出建議參數(不可控)與可控參數的非線性迴歸模型，藉由此可控參數與建議調整參數(不可控)之關係模型，再依據該建議參數(不可控)調整數值為目標執行基因演算法(Genetic algorithm)找出四個可控參數的調整量；其中需先將訓練資料中各參數做最大-最小正規化(Min-Max normalization)再進行模型訓練，提高模型準確度。本研究也針對設備的運行狀態進行監測，用以及早發現不正常運作及設備異常的狀況發生，利用專家規則(Expert rules)的判斷來解決此項問題；以t-檢定(t-test)來計算各參數之非可靠度，並將規則內的參數之非可靠度帶入專家故障規則中，分別計算每一條規則的可靠度，當該規則可靠度低於百分之50時，則會給出設備需保養或維運的標的，由此判斷屬於何種設備故障。本研究開發了使用者介面，以利使用者設定相關分析參數及資料庫的連接，將冰機效率分擬合結果、水塔效率曲線、故障可靠度及指定參數當前的狀態視覺化，並顯示分析之可控參數調整的值及故障分析需改善的設備。 ;This paper focuses on the energy-saving maintenance of the chiller unit in the HVAC system. The chiller machine is a very power-consuming device. The efficiency improvement and parameter optimization of such large-scale equipment often require the experience and knowledge of professionals to achieve. This study attempts to analyze and adjust the uncontrollable characteristic parameters of the chiller machine based on the operating data of the chiller machine, and find out the failure factors of the chiller machine if it cannot be adjusted. It is hoped that this study will contribute to the maintenance of chiller machine equipment. This study trains a nonlinear regression model using training data, with the variables being parameter data collected by sensors. In terms of model design, instead of directly combining the physical properties of the chiller, the study uses a black-box method to develop a data-driven research mode. As not all parameters are controllable, four are set as controllable parameters. The recommendations for adjusting parameters given by the system during analysis and diagnosis are not limited to controllable parameters and may include other non-controllable parameters. Therefore, this study further investigates this issue. When the system suggests adjusting uncontrollable parameters, a non-linear regression model for suggested and controllable parameters is established based on training data and the least square algorithm to create a non-linear relationship between controllable and suggested parameters. Genetic Algorithm is then used to find the adjustments for four controllable parameters according to the suggested parameter. Before model training, the training data is normalized to improve the accuracy of the model by min-max normalization. This study also monitors equipment operating status to detect abnormal operations or equipment malfunction as early as possible, and expert rules are established to address this issue. Fault rules are defined based on past experiences, and t-tests are used to detect whether each specified parameter is abnormal. The selected parameters are then incorporated into the expert fault rules, and the reliability of each rule is calculated. If the reliability of a rule is lower than fifty percent, the target equipment is identified as needing maintenance or repair, and the type of equipment malfunction is determined. This study developed a user interface for users to set relevant analysis parameters and connect to the database. The analysis results are visualized, including the efficiency analysis of the chiller, the efficiency curve of the cooling tower, the fault reliability, and the current status of the specified parameters. The suggested values for controllable parameter adjustments and the equipment that needs improvement for fault analysis are also displayed.
Appears in Collections:	[Graduate Institute of Electrical Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	20	View/Open

社群 sharing

Loading...