Chung Cheng University Institutional Repository
Not a member yet
889 research outputs found
Sort by
基於行動邊緣計算及用戶體驗品質之行動網路直播視訊應用的下行無線資源配置機制;QoE-aware Downlink Resource Allocation Scheme for Live Streaming in Mobile Edge Computing
[[abstract]]近年來,行動網路的數據傳輸量持續增長,為了負載更大量的行動網路頻寬需求,第五代行動網路的各項標準陸續制定中,歐洲電信標準協會(ETSI)基於5G行動網路的演進,提出了行動邊緣計算(Mobile Edge Computing, MEC)。行動邊緣計算將運算與儲存資源佈建在行動網路的邊緣,讓用戶裝置(UE)可以就近使用雲端運算與儲存資源,以降低核心網路日益增加的負載。行動邊緣計算具有低延遲、鄰近性、高頻寬、即時取得無線網路資訊及適地性服務等特色,並能提升用戶裝置的體驗品質(QoE)。視訊直播比起其他行動網路的應用需要更好的體驗品質,為了提升觀看視訊直播的用戶裝置的體驗品質,本論文提出了下行無線資源排程方法,利用行動邊緣計算可以即時取得無線網路資訊的特性,即時取得用戶裝置的傳輸速率及正在觀看的直播的MPEG-DASH串流資訊,使用QoE公式來量化用戶裝置的體驗品質,估算每個用戶裝置分配資源後的體驗品質,將Max-Min排程機制結合用戶裝置的體驗品質來進行無線資源分配,進而提升用戶裝置的體驗品質。模擬結果顯示,我們提出的方法可以提升用戶裝置的體驗品質,並有不錯的公平性。
In recent years, mobile data traffic grows continuously. To load demand for more bandwidth in mobile network, the standards of the 5G mobile network are being developed. Based on the evolution of 5G mobile network, ETSI proposed mobile edge computing (MEC). MEC brings computing and storage resources to the edge of mobile network, and enables UE to use cloud computing and storage resources closely to reduce the increasing load on the core network. MEC is characterized by low latency, proximity, high bandwidth, real-time insight into radio network information and location awareness, and improve UE’s quality of experience.Live streaming need better quality of experience (QoE) than other mobile applications. In this paper, we proposed a downlink resource allocation scheme to provide better quality of experience for the UEs who watch live streaming. We used the characteristic of MEC, real-time insight into radio network information, to receive real-time data rate of UE and MPEG-DASH live streaming information. Using the real-time information and QoE function quantify the quality of experience, which is calculated after each UE is allocated resource. The resource allocation scheme we proposed combines Max-Min scheduling with UE’s QoE, to provide better quality of experience to UE. Simulation results show that the method we propose can improve UE’s QoE and has good fairness
雙端固定式壓電石英音叉加速規之研製;Fabrication of Double-End Tuning Fork piezoelectric Quartz Accelerometers
[[abstract]]本篇論文欲發展出一套能使用於現今工具機高速主軸上,希望能夠做到電路設計簡單化、高靈敏、低成本的”雙端固定石英音叉加速規”。由於應用上是以工具機的條件作為考量,除?可靠?及能在??環境下運作之外,亦須針對應用在某些特殊規格上將性能優化。因此在高溫環境下仍要保持一定穩定度感測會非常重要。有鑑於此,根據石英切角材料資訊:Z-cut石英振盪器在20℃-40℃溫?區間的自然頻?約只有小於10ppm 的變化?具有極高溫?穩定性,而此種材料設定上與我們所需要加速規條件符合。並且選用石英的好處是石英音叉樑感受到軸向應?而使振動頻?產生漂移(drift),由此頻?漂移?可預測加速?值 (假如加速?與頻?漂移?的線性比?關係已建?起?)。假如工作頻?是 32 kHz,它的解析?可達0.01 Hz;相較於CMOS-MEMS因為殘留應力關係以及容易受溫度影響問題,使用石英會較為理想。 在雙端固定石英音叉(Double-Ended quartz Tuning Fork)製程上,我們會提出多種可行製程方式搭配新電極設計,將會從晶圓規格、微影(Lithography)參數、蝕刻(Etch)配方、電極設計與製程,搭配半導體製程實現我們加速規元件來因應台灣目前工具機上對於感測器日益壯大的需求,目前依照工具機高頻率段上的需要,我們的元件共振頻率能夠成功被激發於22 kHz至23 kHz的區間上,且未封裝情況下具有不錯重覆性。於現今國內廠商無自主開發加速規來看,這將會是解決於工具機上加速規高昂的成本問題,對於來達到提升未來機械4.0(生產力4.0)概念上,並且在製程方面上窒礙難行之處以及找出解決方案同時,實現這系統下作為應用資料蒐集的感測器。
This study aims to develop a simple and highly sensitive “double-ended quartz tuning fork accelerometer” with simple circuit design, which can be applied to machine tool high-speed spindles. The application considerations are based on the conditions of machine tools. Thus, the accelerometer must be durable and able to work in an adverse environment. Moreover, its functions must be optimized for applications with certain requirements. Thus, a very important feature is that the sensitivity of the accelerometer must remain stable in a high temperature environment. The characteristics of the quartz material meet the conditions required for the accelerometer. The variation in natural frequency of the z-cut quartz crystal oscillator under the temperature of 20℃ ~ 40℃ is less than10ppm and thus the stability under different temperatures is very high. Thus, quartz is chosen as the material for this study. And another advantage of using quartz is that applying axial stress to the quartz tuning fork beam can result in vibration frequency drift, which can be used to predict acceleration (under the premise that there is already a linear relationship between acceleration and frequency drift constructed). If the operating frequency is 32 kHz, the resolution can reach 0.01 Hz. In comparison, CMOS-MEMS is not as ideal as quartz here for its residual stress and the characteristic of being easily influenced by temperature. Regarding the manufacturing process of the double-ended quartz tuning fork, this study proposes several feasible methods combined with new electrode designs. In order to meet the increasing demands for sensors for machine tools in Taiwan, our accelerometer component is developed with considerations of wafer specifications, lithography parameters, etch formulas, electrode designs, and manufacturing processes for semiconductors. Currently, based on the high-frequency band requirements of machine tools, our component’s resonant frequency can be successfully excited within the interval between 22 kHz and 23 kHz. And in the case without packaging, the repetitiveness is fairly good. Nowadays, there is no company in Taiwan developing accelerometers on its own. Therefore, this study can provide a solution to the issue of high costs of machine tool accelerometers. It also helps to achieve Mechanics 4.0 (Industry 4.0) concept in the future as well as to resolve the difficulties in the aspect of the manufacturing process, so that in this system, the sensors can be applied to collect data
間接式切削力與刀具壽命估測技術之研發;Technique Developments of Estimations of Cutting Forces and Tool Remaining Useful Life
[[abstract]]本論文以振動訊號量測分析發展兩項工具機關鍵技術;第一項是間接式切削力估測技術;第二項是間接式刀具剩餘壽命估測技術。第一項切削力估測技術,首先使用實驗模態分析法(Experimental modal analysis, EMA)獲得刀尖點-主軸之轉移函數,並於銑削過程中以黏貼至主軸上端加速規訊號與刀尖點-主軸之轉移函數進行切削力估測。實驗結果顯示,以本研究所提間接式估測技術與直接式動力計量測之切削力在低轉速下一致,然而隨著主軸轉速提升誤差增大,例如6000 rpm以上。第二項刀具剩餘壽命估測技術,提出以提取振動訊號特徵間接檢測刀具磨耗。使用逆轉主成分分析(Reverse Principal Components, RPCA)方法,從各式振動量、切削力特徵中,篩選對於刀具磨耗診斷有效性與貢獻度大之特徵,接著以自組織映射圖估測刀具磨耗與刀具剩餘壽命。實驗結果顯示本研究所提出之技術相較於直接以照片量測相比,誤差值小於10%以下相當精確。
Two techniques regarding of machining based on the vibration signal are proposed. The first one is an indirect cutting force estimation and the second is an estimation of cutting tool remaining life .For cutting force estimation, a frequency transfer function between the force acting on the cutting tool tip and the acceleration on the spindle is determined firstly using experimental modal analysis (EMA). With this transfer function, the cutting force is then obtained during the machining by measuring the spindle acceleration. Experimental results show that the cutting force determined using the proposed indirection measurement agrees well with those measured using the dynamometer at low spindle speeds; however, the discrepancy increases at high spindle speed, e.g. above 6000 RPM.For the tool remaining life estimation, an indirect detection of tool wear using features extracted from vibration signal is proposed. The effectiveness and the contribution in tool wear estimation from various features extracted from the vibration and cutting force are compared using the method of Reverse Principal Components (RPCA). Then the self-organizing maps (SOM) is adopted for estimating the tool wear, and furthermore the tool useful remaining life (RUL). Experimental results show that the discrepancy in the tool wear estimation using the proposed method is less than 10% as compared with that measured directly by photos
使用預測方法達成即時性五軸工具機碰撞偵測之研究;A Research of Achieving Five-axis CNC Machine Tool Real Time Collision Detection with Predictive Methods
[[abstract]]五軸CNC工具機造價高昂,因此防止撞機的技術就相當重要。在工具機防碰撞的領域中,大多數的應用方式都是藉由預讀加工的NC code(G-code)得到加工的路徑,再以各式各樣的碰撞偵測演算法計算刀具與各軸機具間的狀態。然而這樣的方式並不是真正的即時性防碰撞,反而比較接近干涉的模擬,並且無法應用於沒有加工碼的情況中。為解決上述問題,本文引入統計學中的預測模型,針對工具機的歷史軌跡進行預測,獲得下一個時間點機台可能的座標位置,提早進行碰撞偵測來迴避碰撞的發生。 本文比較了數種預測模型並加以改進,修改後的SAIGM預測模型針對Bezier曲線能達到平均誤差0.006微米,最大誤差0.521微米的預測結果。利用本文改良後的SAIGM預測模型,可以使用儲存的工具機軌跡資料來預測6個取樣時間後的機台位置。 在模型建構方面,本文使用(Constructive Solid Geometry)實體構造幾何模型來建構機台模型,而非三角網格,來降低測試的物件數量,進而使分離軸定理的碰撞偵測時間低於0.1ms,平均偵測時間約為44μs。最後以OpenGL呈現電腦動畫做驗證。 本文目的在於發展一套應用於五軸CNC工具機之即時性防碰撞偵測技術,並且使其能應用於實際的五軸工具機之中。本文提出融合預測方法與碰撞偵測的程式架構在模擬中皆能在1ms內完成,可應用於實際的五軸工具機中。文末針對進給率較大的路徑進行分析,在物件數量極低的情況下也能在1ms內完成整個程式的計算。
Five-axis CNC machine tools are expensive but necessary to today’s industry. To avoid the machine tools collision or accident happening, collision detection technology has been much more important than before. Those machine tools collision detection structures were often done by pre-reading the command or interpolation points from the user. Once they got those pre-reading contours, they will be applied to a variety of collision detection methods. However, those structures are not real-time, more like simulation of interference. On the other hand, they can’t apply to the situation without G-codes. To achieve real-time five-axis CNC machine tools collision detection, we will develop a new structure of machine tools collision detection by utilizing encoders’ data and forecast model. By forecasting the position of machine tools, we will able to avoid collision happening when the SAT(Separating-Axis-Theorem) finish. In this research, we compare several kinds of prediction models and improve the SAIGM(self-adaptive intelligence grey predictive model) forecast model. Forecasting Bezier curve contour with improved SAIGM, the average error is 0.006μm and the maximum error is 0.521μm. The improved SAIGM forecast model can be used to forecast up to 6 sampling times. CSG(Constructive Solid Geometry) method is utilized to represent the five-axis CNC machine tools. The computation time of SAT with limited models is lower than 0.1ms, the average computation time is 44μs. Simulation results will show with OpenGL in the end of program. The computation time of this real-time five-axis CNC machine tools collision detection algorithm are totally below 1ms, which means it could be possible installed in real industry. Even with a higher feedrate, as long as the CSG model’s number is small, the computation time will still below 1ms
DLP積層製造物件列印之最佳建構方向;
[[abstract]]隨著積層製造(Additive Manufacturing)技術的逐漸成熟,開始被大量運用於製造領域當中。其優勢相對於其他工法無須複雜的生產程序,當產品出現錯誤時能迅速修改,能有效的減少產品的設計週期。對設計者而言,在模型的設計與修改上更為便利,來因應各領域逐漸要求快速開發成型之需求。 然而在實際的製造上,積層製造的物件與原先設計的CAD模型間存在著一定的精度誤差。主要是積層製造上的限制,使得物件在層層堆積後在表面產生的階梯效應,造成表面、體積上的誤差。為了降低所產生之誤差,本文透過對物件的旋轉,改變物件模型的切層、列印方向。得到該列印方向,在相同層厚的限制下,能取得擁有最佳精度的物件,並透過歸納的方式得到適用於大多數物件的最佳建構方向簡易判斷。 本文研究重點在建立一個適用於DLP(Digital Light Processing)製造形式的粗糙度模型。所建立模型之絕對平均誤差約為0.94um,約為粗糙度最大值的百分之四左右。配合物件旋轉後各個面所對應的粗糙度進行計算,從中得到最佳建構方向,並歸納出其方向的共同指標。將測試物件帶入粗糙度模型進行旋轉計算,在不計算包含支撐面的情況下,四種測試物件之最佳方向皆有共同之特性。即該方向上,朝上水平面面積為最大值者,為最佳建構方向。關鍵詞:積層製造、DLP、最佳方向、粗糙度模型
With the gradual maturity of Additive Manufacturing technology, what started to be used in the manufacturing. The advantage of the production process is not complicated compared to other methods. It can be quickly modified when the product goes wrong, which can effectively shorten the product design cycle. For the designer, it is more convenient to design and modify the model to meet the needs of rapid development of molding in various fields.However, in actual manufacturing, there is a precision error between the fabri- -cated article and the originally designed CAD model. Mainly due to the limitation of Additive Manufacturing System technology, the step effect generated on the surface after the layers are stacked, causing errors in surface and volume. In order to reduce the error generated, this study changes the direction of slicing and printing of the mo- -del by rotating the object. Obtaining the printing direction, under the limitation of the same layer thickness, the object with the best precision can be found, and got the sim- -ple judgment of optimal direction suitable for most objects by the induction method.The purpose of this study is to establish a roughness model suitable for DLP (Di- -gital Light Processing) manufacturing of the 3D printer.The absolute error of model is about 0.94um. That is calculated by calcu- -lating the roughness corresponding to each surface after the object is rotated, the op- -timal direction is obtained from it, and be got the common indicators of the direction by the induction method. The test object is brought into the roughness model for rota- -tion calculation, and the optimal direction of the four test objects has the same cha- -racteristics without calculating the support surface. The direction of upperside of horizontal surface area is maximum, which is optimal direction.Keyword: Additive Manufacturing、DLP、Optimal direction、Roughness mode
流道設計對全釩氧化還原液流電池性能影響之研究;Study on the Influence of Flow Channel Design on the Performance of All-Vanadium Redox Flow Batteries
[[abstract]]全釩氧化還原液?電池(All-Vanadium Redox Flow Battery,VRFB)為近年來新興的儲能系統,適合應用於再生能源,例如與太陽光電發電與風力發電結合,以達到儲能與平衡負載之用。全釩電池具有靈活的設計、響應時間快且循環壽命長等優點。而在實用上,全釩電池的反應面積較大,故需考量到電解液在反應區內所形成之濃度過電位與其分配均勻性等影響,因此本研究著重於反應區的流道設計與分配流道之設計,以找出對電池堆性能提升的解決方案。在本論文中,設計了數種不同形式的流道,並對於這些設計的速度分佈、流量分佈、壓力分佈和等效電阻進行模擬與實驗。在實驗上,我們設計了具有可替換石墨板的單電池,以研究流道設計對VRFB效率的影響。在電池堆中,電解液流動方向的長度為20 cm,以代表實際應用中的反應區域。石墨板上的流道設計為指叉式的設計,並且有數種流道寬度與肋條寬度的組合。其結果表明,流道寬度為5 mm、肋條寬度為10 mm且流道長度為60 mm的設計會提供最佳的性能。
All vanadium redox flow batteries (VRFBs) store energy by the redox reaction of vanadium ions with different oxidation states. They have advantages of quick switch between charging and discharging processes, long cycle life, flexible design on power and energy capacity. In practical applications, the active area of a VRFB is usually thousands of square centimeters. Accordingly, it is a challenge to improve the uniformities of redox reaction and electrolyte distribution within the VRFB. The former can be improved by flow channel design in the reaction area to deliver high concentration electrolyte toward the outlet area to reduce concentration overpotential; the latter can be improved by flow distributor design between the active area and manifold to reduce the lateral electrolyte concentration difference. In this study, different flow channels are designed for simulation. The velocity distribution, flow distribution, pressure distribution and equivalent resistance are simulated and investigated. A single cell with interchangeable graphite plates is designed to study the effect of flow channel design on the efficiencies of a VRFB. The length of flow direction is 20 centimeter to represent the active area in practical application. The flow channel on the graphite plate is designed as interdigitated, and there are a variety of channel width and rib width combination. Results suggest that the design with a channel width in 5 mm, a rib width in 10 mm, and a channel length in 60 mm provides the best performance
半線性熱方程式軸對稱解的有限差分近似;Numerical blow-up behavior for axisymmetric solutions of semilinear heat equation
[[abstract]]在此篇論文中,我們研究半線性熱方程式軸對稱解的有限差分近似。首先,我們提出了一個數值計算法來重現爆炸解的L^p-norm行為,並討論了此與真實解之間的關聯性;再來提出了計算數值爆炸速率的方法,並驗證其收斂性。另外關於爆炸集合,我們也提出了如何來偵測可能爆炸點的計算法。最後,我們報告一些計算結果