Chung Cheng University Institutional Repository
Not a member yet
889 research outputs found
Sort by
基於機械手臂之咖啡拉花案例研究;Case study of latte art with robot
No full text[[abstract]]近年來，機械手臂在產業自動化的應用已經相當廣泛，因為各種產業分布的不同，以及各產業對於機械手臂的需求也有差異，已往機械手臂比較常應用在工業領域上，以取代人力來做一些高危險性或者是會消耗較多時間的工作，而現在機械手臂的應用已經不再侷限於工業領域，將其應用在現實生活中的例子也越來越多，比較有名的創新應用有機械手臂自動泡咖啡，並且也有公開在國際展會上做發表。 也因為如此，我們想到了一個繼自動泡咖啡後的接續應用「咖啡拉花」，我們提出了手臂拉花這個創新應用，並且將它給實作出來，而在本論中我們實作了很多方法來實現這個目的，其中包括硬體設計、逆向運動學、奇異點避免、碰撞偵測，且因為我們的實驗環境是在實際的餐飲營用場所中，所以其他像是實驗環境保護等等因素我們在實作中也皆有考量進去，讓我們這個創新想法應用在現實生活中的可能性大大增加。
In recent years. The application of robotic arms in industrial automation has been quite extensive. Because of the different distribution of various industries. There is also a difference in the demand for robotic arms in various industries.In the past, mechanical arms were often used in industrial fields. To replace the manpower to do some high-risk or work that will consume more time. The application of the robot arm is no longer limited to the industrial field. There are more and more examples of applying it to real life.Some of the more famous innovations are robotic automatic coffee brewing and also publicly published at international exhibitions. We thought of a continuation application "coffee pull" after automatic coffee brewing. We proposed the innovative application of arm pull and made it happen. In this theory, we have done a lot of ways to achieve this goal. These include hardware design, inverse kinematics, singular point avoidance, collision detection. Because our experimental environment is in the actual food and beverage camp. Therefore, other factors such as experimental environmental protection and other factors are also considered in the implementation. The possibilities for applying this innovative idea to real life have greatly increased
基於ICE流程與霧運算架構於P2P即時影音通訊之負載平衡方法;Method of Load Balancing for Peer-to-Peer Real-time Multimedia Communication Based on Interactive Connectivity Establishment and Fog Computing Architecture
No full text[[abstract]]有鑑於人們對網路的依賴性與日俱增，且使用者對網路安全的觀念也不斷進步，使用NAT的環境也因此增加，再加上NAT類型不斷推陳出新，使得NAT穿越日益困難。IETF針對多媒體傳輸提出ICE作為防火牆穿越的方法，但因上述原因，導致使用者無法完全透過STUN伺服器以對等式網路建立連線，使得架構中的TURN伺服器負載量上升，過去雖然有相關論文欲解決此問題，但其目標大多著重於提升以STUN伺服器穿越NAT的成功率，因此本論文基於ICE協定，以霧運算架構提出另一負載平衡解決方法，並以SVM模型所得的負載值作為選擇TURN伺服器的依據。我們在論文最後對伺服器最大負載量、不同的分析模型，以及架構中新增的流程所增加的延遲時間進行討論，並得到可以在不改變ICE協定，以及僅增加少量延遲時間的狀況下達成負載平衡。
In view of the increasing dependence on the Internet and the improvement of users' concept of network security, the environment using NAT has also increased. In addition, the NAT type has been continuously changed, making NAT traversal increasingly difficult. The IETF proposed ICE as a method for firewall traversal for multimedia transmission. However, due to the above reasons, the user could not completely connect with the peer-to-peer network through the STUN server, so that the load of the TURN server increased. Some researchers proposed a few methods to solve this problem, but their goal is mostly focused on improving the success rate of the STUN server traversing NAT. Therefore, based on the ICE protocol, we proposed another load balancing solution based on the fog computing architecture. In the solution, we used the load value obtained by the SVM model as the basis for selecting the TURN server. In experiments, we showed the maximum load on the TURN server and compared the accuracy of the different analysis models. We also discussed the increased latency of the new processes in the architecture. At the end of this paper, we achieved the load balancing without changing the ICE protocol and added only a small amount of latency
擴增休眠檔縮短x86快速開機後反應時間;Fast Booting Technique for Reducing After x86 Booted Response Time by Extended Hibernation File
No full text[[abstract]]Swap Before Hibernate技術在進入休眠之前，會將各種程式在記憶體中的分頁換出到置換空間中或是檔案系統中，使記憶體內容大小縮減，而讓製作出來的休眠檔也變小，讓系統可以在短時間內讀完休眠檔進而縮短回復時間，藉此優化系統開機速度。但開機完後使用者的程式的working set可能不在休眠檔中，因此作業系統會從置換空間將所欠缺的分頁讀入，導致程式反應時間感受變長。本論文先對開機後所要立即啟動的應用程式進行優化流程，標記這個程式的working set，並將working set存放於休眠檔中，相較於開機後以需求分頁的方式讀入working set。存放於休眠檔中的working set是以循序讀取的方式讀入而需求分頁通常是隨機讀取。對非揮發性儲存裝置而言循序讀取的速度高於隨機讀取，因此可以縮短系統開機到應用程式流暢使用的時間。
The technique, Swap Before Hibernate, will swap out the paging of various programs in the memory into the swap space or the file system before going to hibernate, so that the size of the memory is reduced, and the generated hiberfile is also smaller. The system can read the hiberfile in a short time and shorten the resume time, thereby optimizing the system boot speed. However, the working set of the user's program may not be in the hiberfile after booting, so the operating system will read the missing page from the replacement space, resulting in a longer reaction time.This paper first optimizes the application to be launched immediately after booting, marks the working set of the program, and stores the working set in the hiberfile, which is read into the working set in the form of demand paging after booting. The working set stored in the hiberfile is read in a sequential read mode and the demand paging is usually read randomly. For non-volatile storage devices, sequential reads are faster than random reads, thus reducing the time it takes for the system to boot from the application
使用基因規劃法整合卷積神經網路多層特徵用於影像分類;A Genetic Programming Approach to Integrate Multilayer CNN Features for Image Classification
No full text[[abstract]]隨著卷積神經網路的廣泛使用，近年來有許多研究開始圍繞在如何有效利用卷積神經網路的多層特徵資訊。在許多領域中，整合從卷積神經網路提取的多層特徵已經被證實是有效的。在過去的研究，常見的特徵整合方式大多為串接以及Fisher vector編碼。本論文提出使用基因規劃法(Genetic Programming)來整合多層特徵，藉由演化機制有系統地迭代並找出更好的多層特徵幫助影像分類。在實驗中我們在不同的影像資料集驗證使用基因規劃法自動找出的多層特徵整合方式有助於影像分類，並分析基因規劃法的特性以及演化的統計趨勢。本論文是少數使用演化式演算法(evolutionary algorithm)應用於多層特徵整合方式的研究。在實驗結果中表明了基因規劃法在自動找出特徵整合方式的潛力，並且啟發未來研究及改進的方向。
Fusing information extracted from multiple layers of a convolutional neural network has been proven effective in several domains. Common fusion techniques include feature concatenation and Fisher embedding. In this thesis, we propose to fuse multilayer information from the perspective of genetic programming (GP). With the evolutionary strategy, we iteratively fuse multilayer information into a better representation in a systematic manner. In the evaluation, we verify the effectiveness of discovered GP-based representations on three image classification datasets, and discuss characteristics of the GP process. This study is one of the few works to fuse multilayer information based on an evolutionary strategy. The reported preliminary results not only demonstrate the potential of the GP fusion scheme, but also inspire future study in several aspects
基於基因演算法的晶格模型蛋白質結構預測實作;Implementation of Genetic Algorithms for Protein Structure Prediction Based on Lattice Models
No full text[[abstract]]蛋白質是生物組成重要的一部分，而蛋白質之功能與其立體結構息息相關。為了瞭解蛋白質功能，使用實驗儀器，Ｘ光結晶繞射或核磁共振來觀測蛋白質結構雖然可行，但會花費過多時間與金錢，也有其功能上的限制。因此使用演算法從給定的氨基酸序列來預測蛋白質結構成為另一個可行的方法。在本篇論文中使用從頭開始方法來實作蛋白質結構預測，結合二維三角晶格模型，三維面心晶格模型，親疏水性晶格模型與能量函數，再搭配強化局部搜索的演化式演算法，最後以MATLAB圖形化界面來顯示出預測結果。此外，從實驗結果中可說明我們所選定的演算法在從較長的氨基酸序列中預測蛋白質結構有較好的表現，而在實作蛋白質結構預測工具中我們提供了選項可供使用者選擇並藉此了解蛋白質結構預測的流程。
Protein is an important part of the biological composition, and the function of the protein is closely related to its three-dimensional structure. In order to understand the function of the protein, the use of experimental instruments, X-ray crystal diffraction or nuclear magnetic resonance to observe the protein structure is feasible, but it will take too much time and money, but also has its functional limitations. The use of algorithms to predict protein structure from a given amino acid sequence is therefore another viable approach. In this thesis, we use the ab initio method to implement protein structure prediction, combined with two-dimensional triangular lattice model, three-dimensional face-centered-cubic lattice model, HP lattice model and energy function, and then with the evolutionary algorithm of enhanced local search, and finally MATLAB graphical interface to show prediction results.In addition, the experimental results show that our chosen algorithm has a good performance in predicting protein structure from longer amino acid sequences. In the implementation of protein structure prediction tools, we provide options for users to choose, and use the tool to understand the process of protein structure prediction
鐵鋁合金摩擦攪拌扣件製程之破裂與疲勞行為分析;Fatigue Analysis of Swept Friction Stir Clinch Joints between Aluminum and Steel Sheets
No full text[[abstract]]本研究的目標是鐵鋁合金摩擦攪拌扣件的疲勞分析，內容包含刀具的改良、剪力及十字扣件的破壞及疲勞行為分析、扣件疲勞壽命預測。首先，採用1.6 mm 6061-T6鋁合金及1.0 mm S45C中碳鋼製作扣件試片。依據扣件結構改良刀具設計，並分析進給深度及再進給深度的效應，確認適銲參數。完成剪力及十字扣件試片的製作，進行疲勞試驗，採用光學及掃描式電子顯微鏡對其破壞與疲勞行為進行分析。結果顯示，剪力試片共有三種破壞模式，但十字試片僅有一種。參考試片破壞模式，本研究使用兩種方法進行壽命預測。首先是結構應力法，此法採用剪力及十字試片的結構應力理論解，配合材料之應力-壽命曲線圖，進行壽命預測。其次是應力強度因子法，此法依據主導疲勞裂縫的路徑，取得裂縫的整體及局部應力強度因子解，結合Paris Law，完成壽命預測。最後，本研究以扣件的結構應力及應力強度因子為參數，成功完成兩種試片的實驗資料疊合(data collapse)，取得結構應力-疲勞壽命曲線及應力強度因子-疲勞壽命曲線，將可對複雜受力狀態的扣件，進行精確的疲勞壽命預測。
The objective of this research is fatigue analysis of friction stir clinching (FSC) between aluminum (Al) and mild steel (Fe) sheets including tool improvement, fracture and fatigue behavior for lap-shear (LP) and cross-tension (CT) specimens, and fatigue life prediction. 1.6 mm 6061-T6 Al sheets and 1.0 mm S45C Fe sheets were used for the FSC process. Tool improvement was conducted based on the hook structure of FSC joints. Appropriate processing parameters were obtained by parameter study for indentation depth and re-indentation depth. After making FSC LP and CT specimens, we then conducted fatigue tests and investigated their failure behavior under quasi-static and cyclic loading conditions through optical and scanning electron. Experimental results indicated that LP specimens have three failure modes, while CT specimens have only one failure mode.According to the failure modes of FSC joints, two life prediction methods were used. First was the structural stress (SS) method, which predicted lives by using the SS solutions of LP and CT specimens, associated with S-N curves of sheet materials. Second was the stress intensity factor (SIF) method, which predicted lives by using the global and local SIFs for the paths of dominant fatigue cracks, associated with the Paris Law. Finally, the SS and SIF solutions of FSC joints were taken as the factors to conduct data collapse for fatigue data of two specimens, and then obtain a SS-N curve and a SIF-N curve to accurately predict fatigue lives of FSC joints under complex loading conditions
應用FANUC人機介面軟體於適應性切削之研究;
No full text[[abstract]]在現今工具機產業朝向智能化發展的趨勢下，台灣工機具廠商也紛紛投入資源開發，以附加價值智慧化功能為例，遠端監控、預警系統、資料擷取、刀具壽命預估、溫升補償、誤差補償、震動抑制、調機與適應性切削等等。一般而言，工具機廠商，以離線的方式透過市售軟體 CAD轉CAM後產生加工程式與保守的加工參數，如固定之進給率進行切削加工，未依據不同切削條件與主軸負載變化而改變進給率，造成切削力時而增大與減小，導致加工效率無法提高與刀具壽命降低。本論文將針對加工機具業所使用較為廣泛之控制器廠商FANUC作為實驗載具並藉由其提供之人機介面軟體，開發客制化人機介面實驗平台，平台依實驗架構不同分為，外掛式與內嵌式，外掛式主要是透過通訊協定傳輸，使得個人電腦(PC)可與CNC控制器連線取得資料或輸出控制，因需搭載PC所以本論文稱之為外掛式；內嵌式主要是透過PC將程式編譯完成後，藉由CompactFlash(CF卡)將執行檔燒入至CNC控制器內部運作，此一方法稱為內嵌式。因切削條件不同，控制器設計需具強健性，本論文最後依據量化回授理論設計一強健控制器，應用於適應性切削進給速度控制，再將其控制器與實驗平台整合，實現線上動態更改進給率之適應性切削控制。
Nowadays,the machine tool industry is trended of intelligent development.The machine tool industry has also invested in resource development in Taiwan,taking some intelligent functions for example,remote monitoring、warning system、servo data capture、tool life estimated、temperature compensation、error compensation、vibration suppression、CNC machine parameter tuning and adaptive feed rate control.Generally,the machine manufacturers use CAD/CAM system to produce the machining program which called G-code and conservative machining parameter like fixed the feed rate.Not based the different of cutting condition and spindle load on changing feed rate then resulting in cutting force sometimes increases and decreases.As a result, machining efficiency can not be improved and tool life will be reduced.This thesis will focas on FANUC which is a controller manufacturer and widely used in machinery industry as a experimental platform. With the HMI programming library which provided by FANUC,developing the customize HMI platform. The platform according to the experimental structure is divided into plug-in and embedded. The plug-in mainly through the protocol,making personal computer(PC) access information or output control with CNC controller,due to the structure need PC so this paper called plug-in. The embedded way is compiled a application by using PC and load the application in to CNC controller through compact flash(CF card). Because the change of cutting condition, needed a robust controller. So this paper used QFT(quantitative feedback theory) to design a robust feed rate controller and also integrated it with experimental platform. The integration realize adaptive feed rate control that can dynamic change feed rate online
應用低頻注入法於永磁同步馬達之無感測器速度控制;Sensorless Speed Control of PMSM Using Low-Frequency Signal Injection
No full text[[abstract]]本論文主要為應用在永久磁鐵同步馬達的無感測器技術開發。永久同步磁鐵馬達分為表面貼磁式和凸極式，因為馬達極性的關係，兩者所使用的無感測器控制技術不同，本文主要是針對表面貼磁式的永久磁鐵同步馬達進行開發。在低速方面，我們應用低頻注入法來做為無感測器控制技術，其優點為不需增加額外硬體設備且可以應用在表面貼磁式馬達，缺點為會有暫態抖動以及由於隨著速度命令增大，反電動勢亦逐漸變大，導致低頻注入法模塊計算錯誤的原因，導致低頻注入法無法應用在高速度命令的情況。在高速度命令使用方面，我們應用參考模型適應法，它是一種自適應性的無感測器控制技術，在中、高速度命令下可以有很好的速度響應結果。而在低頻注入法、參考模型法兩種無感測器技術切換方面，我們藉由調整其中的控制參數來改善兩者的可控制速度命令範圍、速度響應，藉以減少切換兩個無感測器技術時所造成的速度不穩、抖動等情形。在模擬方面，我們使用MATLAB Simulink來驗證永久同步磁鐵速度閉迴路控制、低頻注入法以及參考模型法的結果;實驗方面，則是使用德州儀器TMS320F28379晶片來實現。
This thesis mainly builds up the permanent magnet synchronous motor sensorless vector control technology. The permanent magnet synchronous motor can be divided into two types: surface-mounted and interior-mounted. Because of the saliency of motor, we have to choose different sensorless control technology for different types of motors. In this thesis, surface-mounted permanent magnet synchronous motor is used for DSP controlled target.In low speed command situation, we utilize low-frequency signal injection (LFSI) method for surface-mounted permanent magnet synchronous motor to perform sensorless control. Because of the back EMF of the motor, this method can only be used in low speed command situation. In high speed command situation, we use model reference adaptive control system (MRAS), which is self-adaptive sensorless control technology and it performs well in high speed command situation.For the stability when changing sensorless control technology for different speed commands, we adjust the control parameters to improve the dynamic response.We use MATLAB Simulink to identify the result, and use TMS320F28379 digital signal processor (DSP) motion control card to carry out the sensorless motor control
中紅外光矽基鍺錫電致吸收光調變器;Mid-infrared GeSn Electro-absorption Optical Modulators on Silicon
No full text[[abstract]]本論文的主要目的是開發可操作在新興中紅外光範圍內的矽基鍺錫電致吸收調變器。該調變器有著低插入損耗、高消光比與製程整合容易等優勢，可用於高速調變。其原理是在鍺錫合金上施加電場以改變吸收係數，藉由使用電致吸收效應達成光訊號調變的能力。本研究所採用之鍺錫合金擁有較小的能隙，可以通過錫濃度改變做調整，將光學調變範圍從近紅外光延伸至中紅外光範圍。本研究製作並且展示正向入射式的中紅外光電致吸收調變器，元件是使用低溫分子束磊晶技術在矽基板上生長鍺錫 p-i-n二極體結構，使用CMOS兼容的工藝製造元件。從光電流量測中，我們得知鍺錫合金在1600nm的光響應度為0.3A/W，並從中萃取材料吸收係數。接著從穿透光實驗中觀察得到，在1600 nm - 2200 nm光頻譜範圍內的穿透光強度隨著電場的控制而有顯著變化，表明了中紅外光範圍的調變機制實現。此外，施加電場下的元件，光吸收邊緣附近的吸收係數表現出明顯的Franz-Keldysh effect，並且根據公式計算出吸收係數變化與調變比例。本文研製的光調變器的工作範圍為2060-2200nm，在工作區間內實現大於2的高吸收比，且擁有小尺寸的優勢使其能整合於矽光系統。這些研究成果證明了能應用在新興中紅外光的鍺錫電致吸收光調變器。
The main objective of this thesis is to develop Si-based GeSn modulator which can operate in the mid infrared range. The modulator has advantages such as low insertion loss, high extinction ratio and easy integration of process, suitable for high-speed modulation. Optical modulation is achieved by using the electro-absorption (EA) effect, in which the absorption coefficient is turned by applying an electric field across the GeSn active material. In addition, GeSn alloys features a smaller energy gap which can be adjusted by the Sn composition, allowing for extending the optical modulation from near-infrared to mid-infrared. In this study, normal-incident mid-infrared GeSn electro-absorption optical modulators are fabricated and characterized. GeSn p-i-n heterostructures were grown on silicon substrates using low-temperature molecular beam epitaxy techniques. The devices are fabricated using CMOS-compatible fabrication processes. The transmission experiments show that the transmitted light intensity in the spectral range of 1600 nm – 2200 nm considerably changes with the application of electric field, demonstrating mid-infrared optical modulation. In addition, the absorption coefficient near the band edges are confirmed to be strongly dependent on the applied electric field, showing the Franz-Keldysh effects. The working range of this optical modulator is determined to be 2060-2200 nm, in which a high absorption ration of >2 is achieved. These results demonstrate new GeSn mid-infrared optical modulators needed for mid-infrared silicon photonics applications
共振腔增強之矽基鍺錫光偵測器與發光二極體;Resonant-cavity-enhanced GeSn Photodetectors and Light-emitters on Silicon
No full text[[abstract]]本文之研究目的為發展新型 IV- IV 族鍺錫合金光偵測器與發光元件,利用低溫分子束磊(MBE)成長於矽基板上,且 IV- IV 族鍺錫合金與 CMOS 製程相容,使用黃光微影製程製做元件便如魚得水。藉由設計共振腔結構,可大大提升響應與發光,用以改善四族半導體材料在先天上發光特性的不足,並能夠有效地將偵測波段涵蓋整個通訊波段(1260-1675 nm)。在光偵測器方面,我們對元件進行電性與光性量測。偏壓在+1V 與-1V 之間可以看到其良好的二極體特性,正逆偏壓比例達到 173,可清楚觀測出其二極體整流現象;在偏壓為-1V 時,能夠看到有無照光之光電流產生,能夠看出元件有著光偵測器之性能;在光響應度量測上,於光源波長 1550 nm 下可達到 0.35 A/W,和未具共振腔之元件相比增加了 30%,偵測範圍也能夠達到 1800 nm,我們預計增加錫的摻雜濃度,提升響應度強度與涵蓋波長。在發光元件,本文所用為正向入射式共振腔結構,利用各層折射率不同產生共振效果以增強元件發光效率,並量測元件的電致發光頻譜以探討共振腔增強之發光特性,且分別於變溫環境與變電流強度下量測,藉以推算出其光學增益,並達成光學損耗減少。未來,將提高錫濃度及改良共振腔設計以實現光學增益,朝著電激發雷射為目標邁進。
The purpose of this paper is to develop a GeSn alloy light-emitting element, which canbe largely enhanced by resonant cavity structure.To improve the insufficiency ofluminescence properties of Group IV semiconductor materials and effectively enhance thedetection band to cover the entire communication band (1260-1675nm).Firstly, this studyuses low-temperature Molecular-beam epitaxy to grow resonant cavity structuralcomponents, and uses the lithography process to reverse Reactive-ion etching (RIE) andwet etching (BOE etching solution) to make p-i-n normal incidence The structure usesresonance effects to improve detection and luminous efficiency.In terms of photodetectors,we perform electrical and optical measurements on components. When the bias voltage is-1V, the charge density is 0.68A / cm2, and the dark current is still too high. In the opticalresponse measurement, it can reach 0.35A / W at the source wavelength of 1550nm. Weexpect to increase the Sn. The doping concentration is used to increase the responsivenessand coverage wavelength.In terms of resonant cavity effect, we use a normal-incidentresonant cavity structure, which uses different refractive indices of each layer to produceresonance effects, and measures the electroluminescence spectrum of the component toobserve the enhanced luminescence characteristics of the resonant cavity and extract therefraction of the active layer. and measured variable temperature and variable currentintensity, respectively, to calculate optical gain. At present, the effect of the resonancecavity enhancement has been significantly manifested in the luminescence spectrum. In thefuture , the Sn concentration will be increased, aiming to excite as a laser.Keywords: GeSn alloy, resonant cavity effective, phtotdetector, emitter