Master of Science in Electronics & Automation

Permanent URI for this collectionhttp://192.248.9.226/handle/123/59

Browse

Now showing 1 - 20 of 31

item: Thesis-Abstract
Noice reduction in control signals of industrial sewing machines using adaptive filtering
(2022) Niranjan KHVC; Edussooriya CUS; Weeraddana C
Control signals of a typical industrial sewing machine are distorted when they are con nected to the controller. Such distortions due to noise appear at the input port of the control signals and they are, in general, nonstationary signals. Furthermore, access to the controller of an industrial sewing machine is restricted. Therefore, such distor tions cannot be attenuated using classical adaptive filters such as Wiener filters. In this dissertation, an adaptive algorithm is developed in order to solve this challenging prob lem. Here, an additive inverse of the distortion is generated and added to the control signals so that the distortion is significantly attenuated. In order to generate the addi tive inverse of the distortion, the Normalized LeasMean Square (NLMS) algorithm is employed as the adaptive algorithm with an external reference signal. In general, the error signal to the filter is the estimation of the signal, However, based on the nature of the adaptive filtering problem, the NLMS algorithm is formulated in a way that, the error signal to the filter is the difference between the noise signal and the estimated noise signal. The experimental results obtained with the control signals of a typical industrial sewing machine confirm that the proposed method effectively attenuates the distortion signal with fast convergence of the NLMS algorithm.
item: Thesis-Abstract
Empirical formulas to estimate the order of 2-D fir fan filters
(2021) Jayathissa RHNS; Edussooriya CUS
Two-dimensional (2-D) finite impulse response (FIR) fan filters belong to a special class of 2-D filters which has the capability of directional filtering. They are used in many applications such as geological and seismological data processing, and array signal processing. In this dissertation, three accurate formulas are proposed to estimate the order of 2-D FIR fan filters designed using the windowing technique in conjunction with the 2-D separable Kaiser window. Maximum passband ripple 𝐴𝑝, minimum stopband attenuation 𝐴𝑎, half of the fan angle θ, passband width B, transition width T and Kaiser window parameter β are used as the main parameters in the derivation of formulas. Here, three estimation formulas are proposed for three different values of transition width, that is T=0.01π rad/sample, 0.05π rad/sample and 0.1π rad/sample by employing two steps. In the first step, a set of filters with different specifications are designed in order to experientially determine the Kaiser Window parameter β and the minimum order of the filter required to satisfy the given specifications. In the second step, the formulas for the Kaiser window parameter β and the minimum order of the filter are empirically derived through multiple linear regression using the data obtained in the first step. In numerical evaluation, statistical means of the absolute error between the estimated and true values of the Kaiser window parameter β and the minimum filter order are calculated. It is found that mean of absolute error of estimated β and true β is less than one. Also, for minimum filter order it is slightly greater. Mean of absolute error of estimated order is varied from 2-14 of true filter order which are 13.85 for 0.01π rad/sample, 3.11 for 0.05π rad/sample and 2.33 for 0.1π rad/sample. The proposed formulas provide very good accuracy for widely employed 2-D fan filter specifications. By using these formulas, a 2-D fan filter can be designed without trial and error to determine the Kaiser window parameter β and the minimum filter order saving significant time in the design process
item: Thesis-Abstract
A Novel multi-dimensional IIR notch filter for attenuating multiple narrowband interferences
(2021) Chandima PKT; Edussooriya CUS
Notch filters are a class of filters that are used to attenuate narrowband interferences. Previously reported finite impulse response (FIR) or infinite impulse response (IIR) notch filters are predominantly limited to one-dimensional (1-D) or two-dimensional (2-D) signals. With emerging multi-dimensional (M-D) signals, such as four-dimensional (4-D) light fields and five-dimensional (5-D) light field videos, design techniques for M-D notch filters, beyond 2-D notch filters, are required to be investigated. In this dissertation, a novel M-D multi notch IIR filter is proposed to attenuate multiple narrowband interferences. The structure of the notch filter is derived by cascading notch filter pairs. Initially, 1-D filter structure is presented and afterwards it is expanded as an M-D filter. The key factor of the proposed notch filter is the flexibility of placing the notches, adding any number of notches in to M-D and bandwidth can be controlled independently. Narrowband interference attenuation has been verified using monochromatic image and a video. Capabilities of the proposed filter are compared against the existing filtering method using unity three-dimensional (3-D) signal. It is clearly noticeable that, the proposed M-D multi notch IIR filter has better performance as well as greater flexibility.
item: Thesis-Abstract
Probability distributions of inter - sample time of event - based sampling encoders
(2021) Wanniarachchi MS; Premaratne MAUK
Bandwidth is the most important resource in telecommunications. Though recent developments have resulted in a significant increase of available bandwidth, the demand for bandwidth continues to follow and new demands are also created with the introduction of new technologies. Internet of Things is one such development that has resulted in increased demand for bandwidth due to the interconnection of smart sensors and actuators to the Internet. Increased demand for limited bandwidth results in congestion which can in tern negatively affect the reliability of the network by causing latency (delay), jitter (delay variation) and data loss (in the form of packet drops). Event based sampling is a strategy of mitigating congestion that does so by reducing network traffic. This is achieved by reducing the effective sampling rate and it is highly successful if the signal exhibits high dependency between samples. Despite numerous empirical studies, no attempt has been made to obtain a probability distribution of the traffic rate of such encoders. This study aims to obtain such a model for a type of event-based sampling known as memory-based event triggering. With a statistical model of the generated traffic, it is possible to get an idea about the network capabilities and effectively mitigate the congestion. Correctness of the statistical model can be verified by the empirical results and it is possible to easily determine the maximum number of sensors for a given network bandwidth with a given quality of service.
item: Thesis-Full-text
Hand activity recognition using a wearable smart glove
(2020) Samarasinghe JNG; Edussooriya C; Rodrigo R
This project is aimed at designing, simulating and constructing a wearable device capable of performing activity recognition to track and monitor activities specific to the manufacturing industry. This was done by designing data capturing glove to capture all necessary signals from the human body and provide necessary filtering to obtain low noise data. This is then passed through suitable pre-processing algorithms to create distinguishing features between activities. The best suited classification and post-processing algorithms were then designed and implemented to classify the captured data in to a specified set of activities. The device was designed with an ESP8266 and a Raspberry Pi coded in C++ and Python respectively. Accelerometer & gyroscope sensors were used to collect data from the human body while a number of classical machine learning algorithms and convolutional neural networks were tested to classify the data. For the activities pointing, wiping, tightening, loosening, picking, holding, pulling, pushing, hammering, walking, holding and walking and turning, the system was capable of classifying the test data with accuracies between 86% - 91%. The null set was classified with an accuracy of 100% with support vector machines with a linear kernel and the post processing algorithm. The same algorithm reached an accuracy of 91.3% for the activity classification while the support vector machine with RBF kernel and post processing algorithm reached an accuracy of 89.7%. The convolutional neural network trained on pre-processed 3D activity images and the post processing algorithm reached an accuracy of 86.2%. The successfully created device will be used to obtain necessary analysis in the manufacturing space to optimize performance of the workers
item: Thesis-Full-text
Vision-based real-time traffic control using artificial neural network on general-purpose embedded hardware
(2020) Zoysa HKG; Munasinghe R
In urban cities, tra c management of intersections is a substantially challenging prob- lem. In appropriate tra c control leads to waste of fuel, time, and productivity of nations. Though the tra c signals are used to control tra c, it often causes problems due to the pre-programmed timing being not appropriate for the actual tra c intensity at the intersection. Tra c intensity determination based on statistical methods only gives the average intensities expected at any given time. However, to control tra c e ectively, the knowledge of real-time tra c intensity is a must-have. In this project, vision-based technology and arti cial intelligence (AI) are used to estimate tra c in real-time and control the tra c in order to reduce the tra c congestion. General -purpose electronic hardware has been used for in-situ image processing based on edge- detection methods. A Neural Network (NN) was trained to infer tra c intensity in each image in real-time using a scale of 1(very low) to 5 (very high). A Trained AI unit, which takes approximately 4 seconds to process each image and estimate tra c inten- sity was tested on the road where it recorded a 90% acceptance rate. In order to control the tra c, a ratio-based method and a reinforcement learning (RL)-based method was used. The performance of these methods are compared with a pre-programmed tra c controller.
item: Thesis-Abstract
A study of lower limb motions using inertial measurement units
Lalitharathna, SWHMTD; Munasinghe, R
Lower limb motions are vital for human to maintain their everyday life. However, ability ofthe lower limb movements can be affected by different problems. Due to many reasons there large numbers of the population live with various lower limb disabilities. For these people, sometimes it is not an easy task to perform their normal daily life activities. On the other hand, as potential solutions, lower limb prosthetics have been proposed for lower limb amputees and lower limb exoskeletons have been proposed for assisting and rehabilitation processes for the lower limb disabled individuals. However, design and development of the control techniques for such bio-robotics devices is not a straight forward task. Often study and analysis of lower limb motions are important for design and development of the control techniques. Among several methods of studying lower limb motions, Inertial Measurement Units (IMU) based methods have been able to gain lots of attention due several advantages over other methods. Moreover, IMUs can be used in control approaches of bio-robotics applications such as prosthetics or exoskeletons as potential input sources. In this context, the objective of this thesis to study about the lower limb motions using IMUs and investigate the potentials of IMUs to be used in control approaches of bio-robotics applications such as lower-limb prosthetics and exoskeletons. First half ofthis thesis focused on analyzing the human lower-limb motions using IMU sensors mounted over the thigh, shank and foot ofsubjects. IMU sensor data were recorded during walk on horizontal floor, stair ascending and stair descending motions. Comprehensive analyses of lower limb motions were conducted based on recorded accelerometer, gyroscope data of IMUs and sensor fused data. Furthermore, signals from the accelerometer which mounted over the foot were used to detect the heel strike event of the lower limb motions. Based on this heel strike recognition, recorded data were segmented and analyze between each gait cycle. The second half of this thesis, mainly attempted on classification/prediction of walking mode (i.e walk vs stair ascending vs stair descending)and continuous estimation ofthe impaired leg’s foot motions using an IMU mounted over the sound leg’s foot to be used in control approaches of bio-robotics applications such as ankle exoskeletons and lower-limb prosthetics. All the proposed methods were experimentally validated and results highlighted the potential use of IMUs in lower limb motion capturing and control approaches of bio-robotics.
item: Thesis-Abstract
Design and simulation of fuzzy inference based multiple PID controllers for 6-dof unmanned underwatter vehicle
Makawita, CD; Munasinghe, R
Unmanned underwater vehicles are currently being utilised for scientific, commercial and military underwater applications. These vehicles require autonomous guidance and control systems in order to perform underwater tasks. Modelling, simulation and control ofthese vehicles are still major active areas ofresearch and development. This thesis explores the design of a control system for a 6-Dof unmanned underwater vehicle. The thesis consists oftwo phases; the first involves the design ofthree single decoupled PID controllers for surge, yaw and depth. Then it is shown that it is not possible to cover the entire range of operations of UUV using only single controller by simulation using MATLAB SIMULINK. The second phase is concerned with the design ofmultiple PID controllers covering the entire range of UUV operation, as well as the fuzzy inference based supervisor design to switch between the different controllers as the operations conditions vary. The design ofthe PID controllers are based on MATLAB PID tuning algorithms which is a robust response time tuning algorithms that allows for faster design process with robust gain values. It is shown that these new tuning methods as well as graphical tuning interface overcome the adhoc and time consuming process offinding the PID gains. Further it is shown that fuzzy gain scheduling using fuzzy inference mechanism is a valid method for controlling a UUV with nonlinear dynamics. It can be concluded that new tools such as MATLAB tuning algorithms and Fuzzy toolbox allows for fast and accurate design of controllers for highly complex systems as well as the viability offuzzy inference multiple controllers as a method for UUV control with desired response characteristics. Finally the author recommends an actual vehicle implementation and testing as future work to be carried out.
item: Thesis-Full-text
Design and implementation of an automatic wire cutting and striping machine for small scale industry
(2019) Kanagalingam S; Edussooriya CUS
The trend in the recent industry is to move towards automation. This is driven by a number of factors such as increasing accuracy and decreasing human errors. This dissertation provides full overview of the development and design of the automated wire cutting and striping machine for small scale industrial application. The proposed system is put into practice in real time Manual approach is currently used to cut and measure wire that takes more time with manpower. The effectiveness and accuracy obtained by manual method is really poor. The specific aim of the automated wire - cutting system is to cut the needed wire length in the required number of parts. By utilizing the developed system, we can achieve low cost cutting with reduced cutting process time. This system is less complex in terms of user friendliness and also portable.
item: Thesis-Full-text
Jvalue based biomass and growth rate estimation of Duckweed
(2019) Wimalaratne DRSK; Premaratne U
Duckweeds are known as Lemnaceae, comes under the family of small aquatic plants which grows forming a mat covering the surface of the water. Worldwide duckweeds are used as an effective wastewater treatment through conventional methods. These natural green plants remove the excess amount of nutrient or pollutants from the water body and maintain sustainable environmental conditions. Spirodela polyrhiza, Lemna minor and Azolla pinnata are some of the most popular duckweeds used in phytoremediation. Depending on the growing environment, these plants has ability to reproduce rapidly. Rapid growth of duckweeds leads to dysfunction of water bodies and caused other problems. Because of that it is important to monitor the growth rate to control the growth and to avoid an excess duckweed. Traditional method of monitoring the growth rate by manually is laborious and time consuming. Automation of growth rate monitoring is important mostly for duckweed cultivation, modeling of waste water stabilization ponds and among researches. Vision based image processing, used here to automate the growth rate monitoring of duckweeds. For that images of three plants were collected by capturing images from a camera once a two days. In this research two methods were used to estimate the green layer of the three plants Spirodela polyrhiza, Lemna minor and Azolla pinnata. Here the biomass estimation of small fronded aquatic plants is performed by identifying the regions with texture using J- value which is homogeneity measure used in JSEG algorithm. To compare the accuracy alternative Green layer extraction (GLE) method was used. The colour appearance of the surface of the three plants depends on light level, material properties, quality of the images and the view point. For each plant, it was done the green layer detection under two methods with three illuminance levels. Results were verified with the ground truth. According to the results, it was calculated and compared the accuracy percentages and error percentages of two methods in different three illuminance levels. The mean accuracy under normal illumination for the proposed JVT method is Spirodela polyrhiza is 85%, for Lemna minor 82.93% and 83.71 % for Azolla pinnata. Furthermore, JVT method is robust enough to deal with different illuminance levels.
item: Thesis-Full-text
Cost savings from enhanced noise reduction based blowdown control for the State Pharmaceuticals Manufacturing Corporation (SPMC)
(2019) Jayasundera PD; Premaratne U
Boilers are widely used in most of the processing industries like Pharmaceuticals Manufacturing, for the heating applications. State Pharmaceuticals Manufacturing Corporation (SPMC) is the one of the largest Pharmaceuticals Manufacturing Plant in Sri Lanka. In Pharmaceuticals Industries Boiler is mainly used for the steam generation. In an industry normally a 4% of heat energy [1] is wasted through Blowdown. For every Boiler there is a defined limits allowable TDS in the Boiler Drum. The Boiler feed water has a certain TDS. The maximum allowable TDS in process Boilers is 3500 ppm. The steam is generated from the Boiler the TDS contained in the Boiler drums starts to increase. Therefore some amount of the high TDS water needs to be removed from the Boiler drum. This removal of high TDS water from the Boiler drum is called blowdown. By doing Boiler blowdown able to maintain the TDS in the Boiler drum to its optimal desired levels. The process of blowdown is that most of the time the blowdown is done by manually. Therefore that extract amount of blowdown required is never done. Many times excess of blowdown is done and many times sub optimal blowdown is done both these are harmful. Excess amount of blowdown contributes to the blowdown loss. Automatic blowdown control system sensors the actual TDS level in the Boiler drum and does the blowdown only when it is required to do so. When the TDS level in the Boiler go beyond desired set point the blowdown valve will opens and brings down the TDS to the desired level. High TDS level in the Boiler drum not only cause scaling within the Boiler drum and on the Boiler tubes. But these scales particles also get carried away with the steam and formed deposits on the downstream equipment and piping. High TDS levels in the Boiler drum also results in moisture carryover which means steam coming out of the Boiler has a high moisture contained and this is detrimental iii to the equipment downstream of the Boiler. It also results poor heat transfer in the process equipment and cause high steam consumption. An automatic blow down control can keep the blow down rate uniformly close to the maximum allowable dissolved solids level, while minimizing blow down and reducing energy losses. The Boiler Blowdown analysis and energy savings analysis has been carried out at SPMC Plant. The objective of this research work is the blowdown analysis in the plant and analyze the annual savings obtained from Automatic Blowdown Control Intervention. The study was revealed that changing from manual blowdown control to automatic blowdown control monitory savings Rs 202873.11 a boiler’s energy use by annually. Purchasing TDS sensor, pneumatic blowdown valve, PLC and related accessories fabrication cost can be a 3 year payback period on the investment.
item: Thesis-Full-text
Detection of elephant intrusion with seismic sensors and machine learning
(2018) Jayathunga SADS; Premaratne MAUK
Human elephant conflict (HEC) is a severe social issue in several Asian countries. A possible approach to prevent HEC is to identify the presence of elephants remotely, thus people can get precautions. This research introduces a method to detect presence of elephants by acquisition of seismic signals generated by their footfalls. A seismic sensor – Geophone was used to convert seismic waves to analog signal and then it is converted to digital domain. Digital signal processing techniques have been used to develop an algorithm that distinguishingly identifies subsequence signals due to elephant footfalls. Developing such an algorithm was a major objective of the research. A novel algorithm has been developed based relative harmonic contents of the transient signal generated by elephant footfalls. Machine learning algorithms have been used to get the intuition and obtain this algorithm. It uses features of transient signal generated by a single footfall; thus a detection result is generated for every individual footfall. This makes real-time detection possible. Data acquisition and recording hardware has been designed. Site recorded data was processed and analyzed offline in MATLAB environment with a laptop computer. Development and testing of the algorithm was done entirely in MATLAB. However algorithm was designed to be implemented with much less computational power in a microcontroller. Therefore the electronic systems which will use this algorithm can be fabricated as portable units and they can be used at HEC affected areas to get elephant intrusion warnings. Algorithms developed with SVM classification and relative harmonics contents could successfully detect elephant footfalls below average time period of 6s; even when high environmental seismic noise is present. This had been lowered to 3s periods when there is less seismic noise. False detection has average periods of 10s or more.
item: Thesis-Full-text
Analysis of the performance of MQTT with shared dictionary compression (SDC) in IOT networks
(2019) Vinyagamany S; Samarawickrama JG
The Internet of Things or IOT is a set of organized computing devices that are provided with exclusive identifiers and the ability to transfer data over a network without requiring human-to-human or human-to-computer interaction. The Internet of Things spreads internet connectivity beyond traditional devices like desktop and laptop computers, smartphones and tablets to a range of devices and everyday things that use embedded technology to connect and interact with the outside environment, all via the Internet. The architecture of IOT will greatly grow in the next few years and there will a big demand in the field of IOT devices performances. IHS forecasts that the IoT market will grow from an installed base of 15.4 billion devices in 2015 to 30.7 billion devices in 2020 and 75.4 billion in 2025 as shown in the Figure 1 - [2]. In order to cope up with the impending needs, we have to improve the current application protocols used in the internet of things. One of the most popular application protocols for IOT would be MQTT - Message Queue Telemetry Transport). As the Internet of Things ' growth explodes, the underlying fundamental protocols are changing. In particular, MQTT, or Message Queue Telemetry Transport, is now the dominant protocol for IoT globally. MQTT is a machine-to-machine (M2M)/"Internet of Things" connectivity protocol. It has been designed as an extremely lightweight messaging protocol packaged as publish / subscribe. MQTT is alright equipped with compression technologies like deflate. However, our goal is to further enhance compression by introducing Shared Dictionary Compression. Shared Dictionary Compression is tool which uses the redundancies in the messages to form dictionaries for frequently occurring key strings. These dictionaries are distributed among the iv devices and for compression and decommission. This could greatly in terms of the compression and hence the bandwidth. However, it must be noted that enabling Shared Dictionary Analysis would require detecting frequency of repeating keywords among the messages. This could induce additional computation on top of MQTT. Moreover, the distribution of dictionary might have adverse effects on the bandwidth. So, we will need to find the right balance between the achievable bandwidth reduction and computation complexity. In order to find the right tradeoff between computational costs and bandwidth reduction. We will need to implement an algorithm to assess the performance and determine the right settings for the SDC to function. This could be called as the adaptive algorithm, as the settings would greatly depend on the dataset used. In terms of our research, we will initially evaluate the Compression Potential of SDC + MQTT. After confirming the potential of SDC, we will evaluate the same with adaptive algorithm. Consequently, using the adaptive algorithm, we will find the right balance between the performance and compression, by evaluating the compression potential and computational costs of SDC-MQTT. Finally, to further optimize, we will need to analysis the ideal data format for SDC-MQTT for fully optimized performance of SDC-MQTT.
item: Thesis-Full-text
FPGA Implementation of EEG classifier using LDA
(2019) Ellawala NM; Thayaparan S
Design and implementation of feature classification in Electroencephalography (EEG) signal processing system on Field Programmable Gate Array (FPGA) hardware platform is presented in this thesis. Today there is a growing demand for medical devices which process EEG signals, for which, it is important to implement the EEG processing system in hardware instead of software. Processing of EEG signals consist of extracting features from EEG signal and then processing those features to classify the signals. As of today, in most of EEG processing systems, classification part is done on software platform even though the feature extraction is done on hardware. In this project, classification is done with Linear Discriminant Analysis (LDA), based on the features extracted using Discrete Wavelet Transform (DWT), for EEG signals obtained through PhysioNet website. The hardware implementation was done on Field Programmable Gate Array (FPGA) platform using SystemVerilog Hardware Description Language (HDL). Final design has minimum resource utilization, hence is able implement on Basys 3 Artix-7 FPGA Trainer Board with the accuracy of 80%. Therefore, it is concluded this design is suitable for developing low cost, marketable products like sleep detectors for automobile divers. Nevertheless, ultimate goal is to design a simple Application Specific Integrated Circuit (ASIC) chip, which can extract features and classify EEG, so that the full system can be implemented on a portable mobile device without using software platform.
item: Thesis-Full-text
A PID feedback control system for utilizing a RF emission tube at the maximum efficiency
(2019) Wijayakoon SB; Samarawickrama J
Thermal emission tubes are expensive electron devices regularly used in numerous applications such as Radio Frequency (RF) amplifiers, medical instruments, etc. Such a thermal tube designated as TH558E is used as a RF amplifier within a 250kW Short Wave transmitter in the Sri Lanka Broadcasting Corporation in Trincomalee. In this RF amplifier circuit, a control scheme is integrated with the fine tuning of RF amplifier’s final stage to maintain the desired power efficiency and the output power. The present control system configured by the transmitter manufacturer shows poor control capabilities for some broadcasting frequencies and, as a consequence power efficiency and life time of the thermal emission tube is significantly reduced. In this work, we propose a control scheme which is based on multiple Proportional Integration Derivation (PID) controllers and H-infinity optimality criterion to overcome the deficiencies of the original control scheme. Here, a new controller is embedded with an optimal automated tuning method. It is tested for fine-tuning of the RF amplifier’s final stage. The PID control gains are found using an algorithm based on Linear Matrix Inequality [LMI] ensuring the stability. The simulation and test results prove that the proposed control architecture is capable of providing the desired performance.
item: Thesis-Full-text
Time optimized smooth trajectory generation for 2DOF and 3DOF redundantly actuated cable suspended parallel robots
(2019) Boralugoda MBLC; Munasinghe R
Cable Suspended Parallel Robots (CSPR) are a type of cable driven parallel manipulators (CDPR) that has recently become popular for large workspace operations. They possess many advantages over common parallel robot architectures. They also possess the disadvantage of limited dynamics in motion due to the inability to exert compression and the constant limited downward force, gravity. Further, the redundancy in actuation in planar and spatial robots of certain footprints makes it challenging to determine the cable tensions and suitable dynamics for trajectories. This thesis introduces an analytical model to circumvent the cable tension determination problem using a concept termed as ‘Feasible Acceleration Diagram’. It then designs a novel methodology to generate time optimized point to point straight line trajectories with smooth dynamics for redundantly actuated 2DOF and 3DOF point-mass cable suspended parallel robots while ensuring positive cable tensions. The procedure of determination of kinematics for the trajectory is explained in detail with a test case for the 3DOF 4 cable scenario. Finally, the results obtained are verified by a simulation followed by a numerical method.
item: Thesis-Full-text
Network implementation petri net model for Laxapana power plant complex
(2019) Ganegoda DT; Premaratne U
Laxapana Hydro Power complex consists of five Reservoirs & five Power stations. These power stations are located along the Kehelgamuwa Oya & the Muskeli Oya. The total capacity of the Laxapana Complex is 354.8Mw and 13 Generators contribute their service to fulfill the service. It does not have a precise method to schedule these Generators. The Rule of thumb methods derived from past experiences is the only methodology which is used to schedule the generators. There is a cascaded system operating from water levels & flow rates of the reservoirs. it is essentially required a special Modeling Technique to optimize as the water level & the flow rates of the reservoirs are unpredictably change time to time. It had been used a generator optimization method via Petri net Software by Engineer Lankanath. The purpose of this research is implementing the system after studying these data. Most of the researches have been based on analyzing the previous data but in this (my) research real time data is used for the requirement. In this case, water flow rates, water levels data is rapidly acquired by the control system. Moreover all the data of generators are gained by the system. It is decided the procedures & quantities which the generators should operate after analyzing all this data and it is monitored whether they work properly. Eventually, Procedures & Preventive Maintenance dates etc. are decided & displayed by the AI after analyzing the data acquired. Because of this it is possible to optimize power with less failure.
item: Thesis-Full-text
CMOS leakage power reduction and data retention
(2019) Udayanga GWGKN; Thayaparan S
As silicon technology scaling, leakage power dissipation has become the most significant component from all CMOS power dissipation mechanisms. Minimum Leakage Vector(MLV) is used as a combinational logic leakage power reduction technique when a system is in standby mode. Compared to MLV, though an excellent leakage power reduction can be achieved with power gating technique it has some drawbacks like higher retention time and system state loss. In this thesis we combine MLV and power gating techniques to achieve more leakage power reduction compared to MLV while mitigating prior mentioned drawbacks of power gating. Instead of full chip power gating, we developed a simple algorithm which runs in linear time to identify the prospective locations for power gating once combination logic is fed with its MLV. The algorithm was implemented in tcl language and run on top of design compiler shell for a synthesized netlist. Flip flops and input ports were modified to feed MLV in standby mode while facilitating for partial power gating within the flops without losing flop state to retain the system state back in active mode. Flop modifications were extended to feed MLV in scan mode also so that scan mode leakage reduction can also be achieved while successful scan shifting carrying out. Our implementations were tested with four selected ISCAS89 benchmarks using fast spice simulations with synopsys XA. We were able to achieve 30%-40% additional leakage power reduction compared to standalone MLV. The measured wake up time was always less than 0.25ns for all benchmarks while with standalone power gating this is more than a nano second or couple of nano seconds . Successful operation in scan mode and state retention of flops after standby mode were also verified. Rough estimate in area increment due to newly added infrastructure was also carried out.
item: Thesis-Full-text
Statistical models for long term network traffic in enterprise networks
(2019) Atugoda AWCK; Premaratne U
With the rapid development of the internet it has converted the world into a global village and now a day we cannot even think of a micro second down time. For an instance, user demand has caused the internet to successfully combined with other networks. This expanded development has caused for huge internet traffic loads and network congestion. For solving this key issue of the networks it is important to predict the traffic peaks in the network. These traffic peak is caused by a large amount of data being requested like in a download. If these traffic peaks are predictable then non critical traffic from another network can be scheduled to avoid peak to reduce the congestion and maximize utilization. This dissertation introduces a method to solve that key issue. Curve fitting technique in Matlab and distributing fittings are used to build statistical models of predicting traffic. Once that identifying some drawbacks through curve fitting methodology it has been rejected and statistical models for long term network traffic in Enterprise network is used as the proposed technique. Pareto distribution, Beta-Prime distribution and Exponential distribution are derived as the statistical models to predict the traffic peak in Enterprise network. The analysis is conducted by looking at the predictability of a peak in terms of level crossing of a given level. According to the available literature there was no such technique for predicting traffic peaks. As per the results curve fitting methodology error is significantly high. Beta-Prime and Exponential distribution are not good statistical models of predicting traffics due to huge error occurred when compared to the actual behavior of the network. But Pareto distribution is the best model of prediction on traffics in the network as it has vey less error when compared to the actual behavior of the network. According to the results Pareto distribution is the best statistical model of predicting traffic peak. Once predicting the traffic peak can be scheduled the large data from other network for maximum utilization and to avoid the traffic congestion.
item: Thesis-Full-text
Trajectory planning for 6-DOF robot manipulator based on offline robot programming approach
(2019) Fernando, MDM; Munasinghe, R
Industrial robot manipulators are highly involved in modern manufacturing industries. Robot programming is the procedure to carry out generating a sequence of robot instruction. Teaching method is highly applied where a teach pendent is used to generate the robot programme by teaching one point at a time. This process tends to consume more time and the accuracy can be varied depends on the application. Several other methods are used to program robot movement nevertheless industrial applications of these systems are still developing. Programming tends to be difficult and restricts the productivity and industrial application. Hence, requirement of flexible programming methods is still challenging for inexpert robot operators. Trajectory planning for a robot system is still a developing area where the accuracy, productivity and high quality on various operations are highly concerned. To address these limitations, off-line programming systems can be used where computer systems with realistic graphics, interfaces and features can be used to plan and program robot motions without using robot hardware. The research is aimed to present methods for finding a better mathematical way of optimized trajectory planning of 6-DOF industrial robot manipulator. Computer Aided Design software systems are used to implement off-line programming technique by developing human robot interface in order to create robot moving sequence and achieve required data for further calculations. Welding process of machine head cover using a 6 DOF robot manipulator is used to demonstrate and evaluate the proposed method. Methods for Point allocation along the robot moving path and data extraction are presented. Inverse kinematic model for the 6 DOF manipulator is developed and implemented in order to get joint space data represented by joint angles. Derived data is studied to analyze the manipulator motion behavior while moving along predefined path via points allocated. Robot path planning and trajectory planning with CAD system involvement as off-line programming technique is analyzed by comparing results in order to evaluate the performance of the proposed method.