Master of Science in Industrial Automation

Permanent URI for this collectionhttp://192.248.9.226/handle/123/57

Browse

Now showing 1 - 20 of 55

item: Thesis-Abstract
Automating the latex fractionation process in the crepe rubber industry
(2022) Athukorala VH; Jayasekara AGBP; Rathnayake HHMP
Sri Lanka is the largest manufacturer and exporter of natural Crepe rubber which is one of the purest forms of natural rubber manufactured. Crepe rubber should be manufactured in clean hygienic and controlled conditions in order to maintain its pale color and hardness in the final product. Since the inception of the Rubber industry in Sri Lanka in 1876, there hasn’t been much of a development in the manufacturing processes. We still rely on the mechanisms that the British introduced in the colonial era. The initial stage of manufacturing is called rubber fractionation, a process that is specific to crepe rubber production which is done to remove the impurities in the natural rubber latex. This process includes dry rubber content measurement, standardization (dilution to the standard) of the latex, addition of sodium bisulphite/metabisulphite and agitation for around 2 hours. These are high laborintensive tasks and when closely observed many inefficiencies and health hazards for those who are involved could be identified. Also, the labor shortages have been a major bottleneck in the crepe rubber manufacturing process. The objectives of the project are increasing the efficiency and output yield by reducing the process time and wastage of latex and also minimizing the human involvement in the fractionation process hence reducing the health hazards faced by the employees involved. To achieve these objectives firstly, a more efficient and reliable dry rubber content measurement method that enables high precision standardization and chemical dosing are proposed as an alternative to the current measurement using the metrolac. Then an automated solution is proposed through a working demonstration of a prototype to perform the current manual tasks of standardization, chemical dosing, agitation and determining the process end. The prototype enabled the entire fractionation process to be performed in much lesser times than the observed processes in the manufacturing facilities with higher accuracies in standardization and chemical dosing. Hence suggested improvements are proved viable to be implemented in the manufacturing facilities which will give the manufacturers a high output yield, reduce rubber wastages and enable compliance with export standards in the final crepe sheets. The ultimate goal of this study is to make a contribution to the development of the Crepe rubber industry in Sri Lanka.
item: Thesis-Abstract
Noise filtering for accuracy improvement of conveyor belt type dynamic weighing systems
(2022) Balasooriya BMAN; Pathirana CD; Karunasinghe N
Conveyor belt-type checkweigher machines are important part of modern production factories. They are used to determine the weight of the different types of products without stopping the products on the weighing conveyor. The main challenge of dynamic weighing system is providing high measurement accuracy, while the checkweigher machine is running at high conveyor belt speeds. In this paper different types of noises such as electronic noises, motor vibration noises and noises due to natural frequencies are analyzed in frequency domain and it is discussed how digital filtering such as Notch filter, FIR lowpass, IIR lowpass are effective to obtain the required accuracy level. Then to achieve further accuracy level, system is modelled using first principles as spring-mass system as well as black box model. Finally, the selected model is used to apply Kalman filter. Then Kalman is selected as the best filtering method and applied to the real system and simulation and experimental results & conclusions are discussed in detailed.
item: Thesis-Abstract
Prediction of critical parameters for automation of kiln process using DNN regression
(2022) Fernando WKCPP; Chandima DP; Jayasekara AGBP
Cement kiln, the most energy consuming unit of a cement factory, carries out the clinker manufacturing process, which must be operational with stable conditions to achieve consistent clinker quality and maximum production rate. In order to maintain smooth and stable conditions inside the rotary kiln system (RKS), some process control parameters should vary within their desired ranges. This is achieved by doing some adjustments to the kiln control variables. In most of the cement plants, this overall control can only be achieved by manual control by operators. The physicochemical and thermochemical reactions of the RKSs are not yet well understood due to their complexity. Therefore, the behavioral patterns inside the kiln cannot be determined exactly by the operators. Sometimes they end up with wrong decisions for control variables, which can cause the RKS to become unstable and cause huge losses to the cement company. Few automation research studies have been conducted for continuous prediction of control variables for kiln process. However, not all of them address the actual inefficiencies that occur in processes, equipment, and the entire system by recognizing kiln behavioral patterns. Therefore, the automation of clinker production processes with proper prediction model is necessary and it helps to increase production, improve product quality, reduce production costs and operator interventions. This research study is to predict critical control variables such as fuel rate, kiln speed and waste gas fan speed for given RKS parameters to maintain desired process condition inside the RKS. The RKS of Siam City Cement Lanka Limited is used as the case study. A regression based DDN model is implemented and trained for the best accuracy by adjusting hyperparameters. Model evaluation is done until obtaining a minimum error. The results of the model validation in real time scenario are also presented and discussed.
item: Thesis-Abstract
Smart glove for recognition of sinhala sign language
(2022) Thalpawila V K O N; Chandima D P
Speech and hearing-impaired people used sign language to communicate with each other. Sign languages are made of gestures. The language consists of different gestures instead of letters or words. The purpose of this research work is to reduce the communication gap between normal people and hearing and speech impaired people. The research incorporates a system comprising of a glove-based mechanism, consisting of sensors to recognize the hand gestures for Sinhala sign language (SSL) alphabet. The solution combines electronics, sensors, embedded systems, machine learning algorithms, and natural language processing. The research based on a data glove with flex sensors that measure finger bending and an Inertial Measurement Unit (IMU) to recognise palm-turning gestures of the alphabet. Further, sample data with eleven independent variables and hundred data samples per gesture was used for the purpose. In the proposed system, data is trained and classified using Random Forest machine learning algorithm. And natural language processing (NLP) is completed using a newly developed Application Programming Interface (API) to make Sinhala words. The results show that the proposed algorithm has a better recognition effect on gestures, and is capable of making words and sentences. The accuracy of the model on the prepared dataset was founded as 99% for the target user with regard to random forest classification. Complete training for all possible combination of letters and preparation of words is necessary to continue NLP. Also, the system can customise as an education platform for sign language learners. Further, the developed smart glove can use separately for any other hand gesture base applications, the developed ML base system can use or customize separately for feature extraction of any smart wearable item, and finally, the newly developed Sinhala API can use separately for any Sinhala sign language base NLP research work.
item: Thesis-Abstract
Design and development of a uniform spray coater for spin coating
(2021) Amal WGG; Chandima DP
Semiconductor coating is a principle technique which is used to fabricate semiconductors. Spin coating is the commonly use technique to coat the semiconductors among the several coating techniques such as spin coating, spray coating, physical vapor deposition (PVD) and chemical vapor deposition (CVD). Spin coating is a process used to deposit uniform thin films to flat substrates by applying a small amount of coating material on the centre of the substrate, which is either spinning at low speed or not spinning at all and the substrate is rotated at high speed in order to spread the coating material by centrifugal force. Although there are few researches done with spin coating technique and its improvements, there are several defects such as limited to flat surfaces, high material wastage and etc. Hence, to overcome from the limitations based on spin coating, this thesis presents designing of a spray coating machine which is better than the existing spin coating technique and to be used in multiple applications.
item: Thesis-Abstract
Path planning of a robotic manipulator for 3D scanning of moving object on a conveyer
(2021) Weerasekara DRBKK; Jayasekara AGBP
Robot manipulator systems are used in an industrial automation system to minimize human e ort or involvement and increase product or service quality. Generally, auto- mated robot systems are used to perform material handling, assembly operations, and quality inspection of the manufacturing system to achieve better performance in the precision, accuracy, and production rate of the continuous operation. In the injection molding industry, the surface quality of the output product is critical for producing a high-quality product. The surface scanning method is a common method for inspecting product surface quality. Most of the existing scanning methods need to identify and place the object in a xed and known position to do quality scanning. In this research, we proposed a method of using a 4 DoF robot manipulator to move the scanner to get a quality surface scanning output for injection-molded products which are moving on a conveyor. Robot manipulator path planning is one of the main objectives of the research. The movement of the robot's end-e ector must change depending on the object's orientation. The object image feature extraction method is used to determine the orientation of the object.The angle value's maximum accuracy ranges from +50 or -50 .The robot's end e ector angle is changed according to the measured orientation of the object. The robot end e ector is required to follow the object without any relative speed on the conveyor and maintain the absolute maximum speed to achieve an ef- fective scanning output.According to experimental results, the optimal conveyor speed is 10cms-1.The speed control system is used to maintain the conveyor speed without any external disturbance and measure the speed, and it is fed to the robot system to maintain the relative speed of the conveyor. The distance between the scanner and the object is measured using an ultrasonic sensor. This sensor feeds the distance to the system, and the distance helps to maintain the path trajectory and takes into account the quality of the scanning output. A portable 3D scanner, the Quanser Kinova 4 DoF robot arm, and the MATLAB Simulink platform are being used to simulate the proposed system.According to the results, the conveyor speed was set at 10 cms-1 and the robot end e ector moved on the trajectory based on the object orientation angle.
item: Thesis-Abstract
Design and development of semi - automatic tire inspection machine
(2021) Perera HAPB; Jayasekara AGBP
Machine learning has become an important and interesting eld when addressing com- plex industrial tasks. In this study, a semi-automatic tire inspection machine for tire retreading industry which uses machine learning techniques is developed. Most consumers tend to retread their tires because retreading is an economical and an eco-friendly method. As a result, retreading industry is now becoming popular in developed countries as well as developing countries. Initial inspection is the most crucial activity in the retreading process because tire defect identi cation is performed in this phase. Failure in identi cation of defects prior to retreading, may cause delamination of a retreaded tire consequently leading to a disastrous accident. There exist many advanced machines for initial tire inspection. Widely used method in the tire retreading industry is nondestructive defect detection based on X-Ray image processing. This method is very expensive and used by brand new tire manufacturing companies as well as tire retreading companies in developed countries. In addition to that, devices with holography and shearography techniques are used to identify defects which map the tire defects using optical means and are known to be extremely expensive. Conventional inspection method is being followed by the Sri Lankan tire retreading industry as well as other developing countries such as India, Bangladesh etc. due to its cost e ectiveness in replace of extremely expensive advanced machinery. The two main tasks performed in conventional inspection method are visual inspection and hammering test. Operator carefully observe the worn tire, identify and mark the defects which could be observed through the naked eye under visual inspection. The identi ed defects can be classi ed as tire punctures, unwanted metal particles, ply damages, bead damages and side wall damages. Hammering test is carried out to identify the defects which are invisible to naked eye such as inner ply separations, ply damages of a tire, inner canvas damage and small air bubbles in tread area etc. Usually the test is performed by hammering all over the tire tread area using a brass rod and listening to the resulted noise di erence by an expertise. An expert human inspector performs both visual inspection and hammering test. This manual inspection is often associated with inaccurate results and undetected defects due to lack of expertise, causing visual fatigue which results in low e ciency and higher amount of labor costs in local tire retreading industry. Therefore, the main objective of this study is to eliminate the expert human resource from the initial tire inspection process and reduce the complexity of the activity. In this research, visual inspection activity is trained to a model and Faster RCNN Inception v2 algorithm is used on TensorFlow platform. Image classi cation and tire defect detection are done with a collection of real-world industrial image data set. These images were captured using four cameras which were having a capacity of 12 mega pixels each. Basically 220 images were trained using a computer. Hammering test activity is trained to a model and YouTube-8M algorithm is used with VGGish feature extractor on TensorFlow platform. The sound signal was captured via a normal USB microphone. The sound signal was analysed using Audacity open source software and fed as the input to train the model. Unwanted metal particles of the worn tires are detected using a metal detector. In addition to defect identi cation mechanisms, defect localisation system is developed using a microcontroller and an encoder. Defects which are identi ed from image processing or sound signal processing or metal detection, location of the defect is identi ed with respect to a reference point of tire. This is very useful for identifying the exact defect location of tire for the operator. From the obtained results it can be concluded that, the above image classi cation and sound signal classi cation models provide results with a higher level of accuracy. Therefore, the expertise labor which is used to perform the initial inspection process could be replaced by a novice employee. Furthermore the unique and ideal structure of this developed machine is associated with low maintenance cost. As a result, small scale companies would be more comfortable with their existing nancial situations when using this semi-automatic tire inspection machine to enhance their throughput of the tire retreading process.
item: Thesis-Abstract
Intelligent railway signaling failure alert system
(2021) Sugandi MAN; Jayasekara B
One of the key factor for the development of the country is the efficiency of the transport system. In Sri Lanka, 20 percent of the passengers travel by the train. The demand for the rail transportation systems have been increased by the increment of rail transportation facilities and the population. Recently rail transport was introduced to Sri Lanka in 1864 to transport goods, but currently popular for the passenger transportation. The railway signaling system is used to cater the traffic issues by utilizing the limited resources such as trains and tracks. The loss due to the failure of the signal system is substantial, as this result in the loss of human lives and human hours in addition to the loss of damages to properties. For the reliable signaling system main input is train detection. Currently Sri Lanka Railway hasn’t got a proper Failure diagnosis and maintenance reporting system. Therefore, it is essential to have a system to ensure that railway signals are reliable and safe. The research is based on an intelligent fault detection method on railway signal color light signals. Before commencing the research, several literature surveys had been done on railway signalling and railway fault diagnosis methods, railway signal maintenance and signal fault detecting mechanism. Moreover attention was placed on other industrial intelligent fault diagnosis methods. A brief go through on signal aspect circuits, circuit relays and relay theories, would help find out the most suitable method for fault detection. To carry on the research many color light signals inputs, outputs were obtained and surveys were done on different signal lamps, signal poles, signal aspect faults and faulty ranges in each red, amber, green circuits were identified. Mainly research was done by a case study, from the three signal aspect pole 263 where located in Colombo- Fort yard. All the data of case study 263, three signal aspect pole’s data was recorded manually and the current and voltage reading fault ranges in each red , amber, green circuit’s electric components were observed. Considering observed data and survey data model is implemented for the case study. The Model is implemented with three input subsystems and six outputs. The proposed model uses a fuzzy system and the model is design by MATLAB Simulink system. Fuzzy system is developed using the “mamdani” technique. The system is tested with actual readings for different faulty modes and justified that model output result and actual readings are accurate. The final develop model shows that the proposed method has the capacity to find faults in each Red, Amber and Green color light circuits.
item: Thesis-Abstract
Enhancing the capabilities of intelligent wheelchair robots in approaching and docking to service scenarios
(2021) Hiroshaan V; Jayasekara AGBP
The world is presently confronted with the issues of aging and disability as a result of accidents and other events. For the above problem, wheelchairs are the evaded partner in the lives of many differently-abled people to support their day-to-day activities. Many academics are focusing on the usage of wheelchairs to discover better robotics solutions. However, the status of the automated powered wheelchairs is not up to the required level of autonomy in the areas such as docking behavior for a specific task. Approaching and docking need correct layout information, and identifying the proper degrees of docking in the environment and then assessing them for safety. A machine learning system trained on various docking-level setups is one viable approach. However, retraining is required for different scenarios such as furniture placements or real-time random changes of the layout. Furthermore, vision data is affected by changes in light conditions. The method for detecting docking surfaces, on the other hand, is dependent on geometric information computed from depth data, which makes it invariant to scene or light changes. As the main aim of this research, a human study was performed to identify the docking behavior of a wheelchair to the table or desk in four different scenarios such as writing, reading, eating, and using a laptop with 3D point cloud data. This research developed a novel method for determining comfortable docking locations based on analyzed ergonomics data. It was gathered from human subjects on actual wheelchair usage. Analyzed data can be applied within a single algorithm to obtain a safe location using 3D point cloud data. While docking with the table, two situations were evaluated. The first is a table with an object on it, while the second is a table with no object on it. If the object is on the table, it will dock based on its location on the table and the availability of open space. If no item is present, it will dock based on the user's desire and the available free space. This wheelchair also has navigation and obstacle avoidance built in to let it travel in a residential setting more independently. From the human study, the optimized distance between wheelchair back end and table were identified for eating, writing, reading and using a laptop as 29 cm, 27.75 cm, 27.25 cm and 40 cm respectively. The optimized height difference between table surface and wheelchair seat for all scenarios were obtained as the same value as 32.25 cm for a particular table (height = 81 cm). Seat height was not dependent on the scenario. Obtained results were applied to the simulation design for above two situations and validated through fourty test cases.
item: Thesis-Abstract
Multiple turbine flow distribution control using intelligent controller in mini hydro power plant
(2021) Deshapriya ULDV; Jayasekara AGBP
Mini hydropower plants have been located at relatively small streams such streams often found upstream of a large river system. With a small catchment and high runoff, those stream parameters such as flow, debris content and water purity are highly dynamic with seasonal intense rainfall. To generate maximum electricity from river available flow, mini hydropower plants are equipped with multiple turbines. Thus, different flow combinations can be used to harvest maximum energy. The Present flow distribution control technique consists of a PID loop and a pre-tuned flow distribution table which allocates flow to each turbine by changing wicket gate opening depending upon available river flow. But as a result of variable stream parameters, water density and penstock friction losses will drastically change and therefore a flow distribution combination given by the existing controller may not be the optimum combination for the given moment. with alternating upstream rainfalls, most mini hydropower plants are not able to reach maximum design capacity even with 100% flow until clear water appears in the river. This research aims to introduce an artificial neuron network based intelligent controller which will measure flow variables and update the flow distribution table frequently ensuring optimum flow distribution and maximum power generation. “Moragaha Oya” mini hydropower plant located in Kandy district, Sri Lanka is selected as the test site for this research. Required additional hardware has been installed for the water quality and rainfall measurements. Developed intelligent controller with data collection and training software has been installed in power plant control computer enabling the update of flow distribution table. The intelligent controller has been trained and performance validation completed. A significant improvement of total output power has been obtained during performance validation, especially during the flow conditions where the existing controller could not obtain the optimum flow condition. the maximum output power increment is recorded as 3% of the previous maximum output power. By using energy harvest data of 47 days and rainfall data of the last 4 years, a 2% increment of annual energy harvest was also predicted during performance validation. Required further developments were also identified. The findings of this research may very important to renewable energy developers to optimize their powerplant yield.
item: Thesis-Full-text
Designing a robust controller to damp sub-synchronous oscillations in power systems
(2020) Gamage CM; Prasad WD
In the power transmission systems, the power transferring capability is limited due to the inductive reactance of the transmission lines. In order to mitigate the inductive effect, some compensation techniques are applied to the transmission lines. One such technique is the series compensation using capacitor banks. Series compensation method is used to improve the system voltage with capacitor banks are connected in series with the power transmission line and it expands the power transferring capability of the line. Although the increase of series compensation improves the power transfer capability and the steady-state and transient stability limit of the power transmission line, it can lead to the generation of some natural frequencies due to the combination of inductor and capacitor (L-C). These frequencies are called as sub-synchronous frequencies which are below the power frequency of the power systems. They can arise sub-synchronous resonance (SSR). The SSR can cause physical damages to the power system equipment unless it is detected and mitigated punctually. Several number of mitigation techniques for different types of power system oscillations have been proposed in literature. But existing mechanisms are not completely damp these oscillations or the mechanisms used to damp these oscillations might be source for any other control situations. Therefore, this is a phenomenon which should understand well and damped these oscillations properly. The intension of the work presented in this thesis is to properly mitigate the undamped power system oscillations which are in the range of sub-synchronous frequencies. This research proposed a robust controller which can damp dominant sub-synchronous resonance. Further, the implemented controller performs well in different operating points. IEEE First Benchmark Model (FBM) is used as the test system and the dynamic phasor representation of the system is used to model the small signal model. The operating points of the test system were generated by changing the series capacitor compensation level of the power transmission line. Finally, this research introduced a robust controller with PID controlling to damp out dominant sub-synchronous oscillations which can perform well under different operating points of the selected power system.
item: Thesis-Full-text
Head and eye operated computer interface for a physically disabled person
(2020) Tuder GD; Jayasekara AGBP
Human computer interaction has the physical and theoretical limit among the human being and the input/output devices of a computer. Studies of severely disabled people have shown that most of disable people can be able to direct head and eye motions. This can be applied to build new human computer interface mechanisms, then it helps to communicate between others otherwise control certain specific devices. The proposed and designed system is a head and eye operated mouse which is expected to be used by a disabled person. The system is divided into main two parts, first one is a Wearable glass module with Arduino Nano, Accelerometer sensor, Eye blink sensor, Bluetooth Module, Voltage regulator and Lithium polymer battery (Transmitting part), Second one is Base station with Processor and Bluetooth module (Receiving part). Developed mouse have accelerometer sensor to identify the head movement of user, to control mouse courser in real time. Infrared sensor to identify the intentional eye blink of the user to activate mouse clicks (single / double) in real time. Developed mouse uses Bluetooth technology to communicate tirelessly with a computer. A usability testing survey was conducted to validate the product, a group of 20 was volunteered in conducting the survey including a disabled person, among the considered group 90% were strongly agreed to the product as a new concept to be used by disabled person. 65% were satisfied with the functionality when compared to the existing mouse. Further 60% of the users were satisfied with the usability in different environments. Finally, 40% were in the position of not satisfied with the wear ability by a disable person by himself. Survey results validate the product that have smooth control, proper perfect movements and good sensitivity as normal mouse operation. Wireless , portable Head and eye controlled mouse will be an easy input device for paralyzed and hand disabled people. The overall operation of device complied with all of the requirement set out in the original design proposal.
item: Thesis-Full-text
Energy management and control of electric bike using hybrid power source
(2020) Moratuwage KI; Karunadasa JP; Jayasekara AGBP
Fuel sources for modern transportation systems are getting pricey and negatively affect the environment which lead to increase the demand for electric vehicles. Energy storage systems in majority of electric and hybrid vehicles are based on battery storage devices. Nevertheless, battery based systems have several issues that caused by high peak power demand which could resolved by high power density batteries .However, high power density batteries are much more expensive which lead to increase the overall cost of the vehicle. Proposed Hybrid system (HESS) which connected to exciting electric bike consist of super capacitor bank ,DC to DC converter, motor controller, Battery bank and BLDC motor.DC to DC converter positioned between supercapacitor bank and battery bank, which pumps required energy to the supercapacitor bank,in order to maintain a greater voltage value than the battery terminal voltage. In most riding occations in a control manner. Only when battery voltage equal to the capacitor bank voltage at continous bulk energy demands, battery connected to the Brush less DC-Motor which Maintain a relatively fixed load profile. Further, regenerative energy generated by braking is also fed to the battery indirectly via capacitor array, thus, battery pack isolated from frequent power demands which caused to reduce number of charge discharge cycles hence, increase the lifetime of the battery. Finally, Test results clearly indicate, this Hybrid energy storage system has enormous benefits compared to Electric bikes such as reduction of overall power consumption of the battery , enhance quick acceleration , increased travelling range per single charge and decrement of per kilometer cost. Further, HESS system more energy efficient, more cost efficient and smooth in running compared to current electrical bikes in the market which makes HESS bike good choice for future higher speed electric bike industry.
item: Thesis-Full-text
An Under-actuated mechanism for an anthropomorphic prosthetic hand
(2020) Chamara RPDD; Gopura R
Upper limbs are very important to the functionality of the human body that enables us to execute activities of daily life (ADL). The human upper limb is complex and difficult to mimic from a robotic manipulator. Over the years the evolution of robotic prosthetic hand development is evident, which brings robotic prosthetic hands closer to the human hand performance. The challenge is to develop a robotic prosthesis with fewer actuators, human-like appearance and convenient interaction for the amputee. This research focuses on developing an under-actuated mechanism for a robotic prosthetic hand while maintaining the anthropometry of the human hand. This under-actuated mechanism primarily generate three grip patterns, namely index finger extended, pinch grip, and power grip. The under-actuated mechanism consists of a clutch mechanism that uses two actuators for controlling flexion and extension of fingers excluding thumb. This reduces control hardware whilst enabling to build a functional under-actuated mechanism incorporating 3D printing for creating lightweight prosthesis. Initial CAD models enabled to check the feasibility of under-actuated robotic hand design and various simulations helped to verify the design. For developed prosthesis hand, the mathematical kinematic model was constructed which was fed into the Matlab software to evaluate effectiveness of motion generation of fingers.
item: Thesis-Full-text
Design of a deep reinforcement learning based optimal PH controller for nitrification bioreactors in aquaponics systems
(2019) De Silva PCP; Jayasekara AGBP
Recent advances in deep reinforcement learning has produced state of the art algorithms. These algorithms have better training stability, convergence and computational performance. In this study a state of the art deep reinforcement learning algorithm is used to implement a self-learning, model free, non-linear controller to control pH of an aquaponic system. Aquaponics is a soil-less farming system where effluent water from a fish tank is used as nutrients for growing plants. Maintaining the pH of an aquaponic system provides the optimal condition for micro-organisms that convert the ammonia rich fish effluent to nitrates, which are easily absorbed by the plants. In order to optimize this conversion process known as nitrification, pH is maintained at optimal conditions within an intermediate setup known as the nitrification bioreactor. The implementation of a deep reinforcement learning based controller is studied in detail and the performance of the deep reinforcement learning based pH controller is evaluated by comparing the performance of a classic PID based controller in an aquaponic system. The results show that DRL based controllers are better suited for control of dynamic stochastic control pH process and is capable of learning complex plant models and tuning itself based on the learnt model. The outcomes of this research can be applied in the design of optimal controllers that learns purely from experience to optimize various industrial processes. This type of controllers is ideal in Industry 4.0 based applications.
item: Thesis-Full-text
Intelligent maintenance management model for critical machines in a solid tire manufacturing factory
(2019) Jayasuriya RPSK; Abeygunawardane SK
High machine reliability is an essential feature for a solid tire manufacturing plant. Most of the machines in those plants should be capable of 24 hour continuous running. Engineering department of a solid tire manufacturing company has a great responsibility to maintain the machine reliability level and to assist a trouble free operation. They conduct preventive maintenance and conditional monitoring regularly to minimize the breakdowns in the machines. However, the current preventive maintenance practice of many engineering teams in solid tire manufacturing plants is a fixed schedule and it does not update with the condition monitoring data. Due to this, sometimes machines are serviced when maintenance is not needed and sometimes they are not serviced, when maintenance is needed. If there is a breakdown due to lack of maintenance and the maintenance team cannot rectify the problem, they have to get assistance from superior levels which may lead to high down times. This work aims to develop an intelligent system to dynamically change preventive maintenance schedule based on machine condition data and breakdown history for critical machines in the Camso Loadstar ETD2 solid tire manufacturing plant. In addition, this work applies artificial intelligence for troubleshooting. The intelligent maintenance management system designed using artificial neural networks and expert system provides a dynamically updating maintenance schedules and troubleshooting assistance. The performance of the designed system is evaluated separately for maintenance scheduling and the trouble shooting assistance. The performance of maintenance scheduling is analyzed using 10 critical machines by comparing predicted results with real achievements. The performance of trouble shooting assistance is evaluated by calculating the maturity level of the established expert system. The results show that the proposed intelligent system is a good solution for the existing issues related to maintenance of the critical machines in solid tire manufacturing plants.
item: Thesis-Full-text
A Neural network based vector control scheme for regenerative converters to use in elevator systems
(2019) Senadheera WS; Hemapala KTMU
Current days, large scale buildings are the major energy consumers in the world. In most of the cases, energy is wasted than using effectively in buildings. Clients always request optimum energy consumption levels when the new buildings are designed. In a conventional elevator system, energy is dissipated as heat in a set of resistors when braking occurs. Using this dissipating power for another useful activity as regenerative power will make the energy usage of a building more efficient. The main modification to be done for the motor drive to collect this regenerative power is to replace the passive rectifier in the drive input side with an active AC/DC converter. Traditionally, these converters are controlled with PI controllers. Though, modern experiments reveal that arrangements of these kinds demonstrate restrictions with their suitability in practical applications. This research explores on mitigating similar limitations by applying a neural network in regulating active front end converters in such systems. Further, it proposes a neural network related switching regulation scheme for bi-directional AC/DC converters to improve the efficiency of extracting regenerative energy in elevator systems. By using this kind of NN controller setup, bi-directional AC/DC converters can achieve the advantages such as quick switching response, simpler structure and better output waveform. Neural network controller’s performance was analysed together with normal vector control stipulations and compared versus traditional vector control arrangements. This establishes that the neural network vector control scheme introduced in this research is more efficient and useful. Even with rapidly changing and power switching converter control arrangements, the NN based vector control mechanism exhibits good performance levels. Following input reference signals which are fluctuating frequently, fulfilling the basic regulating requirements for faulty power utilities and enduring of unstable situations in power regeneration system
item: Thesis-Full-text
Design and development of an interactive robotic conversational companion for elderly people
(2019) Manuhara GWM; Jayasekara AGBP
The ageing of population is rapidly accelerating worldwide and as a result countries are facing social and economic challenges. Hence, the majority of the elderly population all around the world is facing difficulties. The loss of ability is typically associated with ageing and the elders require special attention in both physical and mental concerns. The requirement of a suitable caretaker becomes very important in caring for an elderly person. A human caretaker would be the ideal solution. But the availability of such genuine resource is a very rare luxury in the modern society. Hence the society and the elderly population are in need of a suitable alternate solution. Introduction of service robots has become a very promising development in addressing problems faced by elderly population in the world. This research work proposes a robotic conversational companion capable of vocal interaction with elderly users in human like dialogues, during service assistance. A Finite State Interaction Module (FSIM) and a regular expression based language identification method have been introduced for facilitating this task. A Knowledge Database (KDB) containing specific data has been designed, implemented and connected with the robot system to enable more meaningful and natural dialogue creation. State transition diagram and event flow diagrams explaining the functionality of the states are presented. The robots performance has been evaluated by user rating. Experimental results including a selected segment of conversation are presented with an analysis including the change of FSIM states. Human user has been asked to interact with the experimental setup and rate the user experience varying from “Very Bad” to “Excellent”. The evaluation results have indicated a high user satisfaction rate close to “Good”, validating the robots capability to interact in a human friendly manner during service assistance.
item: Thesis-Full-text
Extended kalman filter and stereoscopic vision based autonomous flying system for quadcopters
Somasiri, JAAS; Chandima, DP; Jayasekara, AGBP
This thesis can be divided into two main modules. First module is implementation of an Ex-tended Kalman filter and introduce into existing flight control algorithm which is used to con-trol multi-rotor unmanned vehicles. Purpose of this implementation is to improve flight per-formance and reliability of the system. Second module is implementation of an obstacle avoid-ance system based on stereo vision and fuzzy logic for same flight control algorithm to avoid crashes and avoid obstacles during navigation. In this thesis Chapter 1 introduce basic modules of this implementations and explain about flight control algorithm and its major components which is used in here. This chapter also explains the theory behind the Extended Kalman Fil-ters, stereo vision systems and fuzzy logic. Chapter 2 described literature survey about existing implementation of Extended Kalman filters on multi-rotor platforms, stereo vision system im-plementations and related obstacle avoidance implementations like artificial potential field and fuzzy logic. First section of chapter 3 focused into implementation details and experimenting results of Extended Kalman filter and also explained how Extended Kalman filter outputs are combined to Attitude and Position controllers of flight control algorithm. Second section of chapter 3 focused into implementation and experimenting results of the stereo vision system. This section explained detail implementation of stereo vision system like stereo camera cali-bration, image rectification, disparity map generation and depth calculation. Mainly OpenCV was used in this implementation. Third section of chapter 3 focused into explained implemen-tation of fuzzy decision-making system. In here described deciding of fuzzy inputs and outputs using depth image, creation of fuzzy inference system, selection of membership functions and combined fuzzy decision-making system with flight control algorithm. Flight testing and ex-perimental results of Extended Kalman filter and obstacle avoidance system were described in chapter 4, both systems were tested on outdoor environments and improvement of the perfor-mance and reliability was discussed in this chapter. Chapter 5 is the final chapter of this thesis and it includes conclusion of the thesis, recommendations and further works.
item: Thesis-Full-text
Robot companion for adaptive home environment
Senarathna, SMSS; Jayasekara, AGBP
Robot companions are being develop to assist humans in domestic environment. Sooner rather than later, robot buddies will be a piece of our typical family's day by day life attempting to help with local errands and dealing with us when required. Notwithstanding, before this turns into a reality, imperative issues should be tended to in the Human-Robot Interaction (HRI) field so as to accomplish social robots equipped for communicating with people also to which people associate with each other. One of the greatest difficulties in this field is to endow robot associates with those social capacities expected to collaborate with a man amid a consistent period. These aptitudes can be upgraded by utilizing robot's own sensors or outer tactile frameworks introduced around the trial condition, so robots could know about the logical data. The fuse of these capacities hopes to enhance the association and the robot colleagues acknowledgment by people. Individuals have desires when initially experiencing a buddy robot, particularly in domestic environments, where the capacity to mingle and impart in a human-like manner and distinguishing the real necessity of the user by utilizing human like knowledge by recognizing user behavior without user to mention it, are crucial highlights to fuse keeping in mind the end goal to accomplish the coveted level of collaboration expected by people. This Research Project was based on to achieve a robot companion which would identify the actual requirement of the user by using human like intelligence by identifying the user behavior without user to mention the requirement by himself and to have the ability to socialize and communicate with the user and with using these two major capabilities control the lighting level, temperature and humidity of the home which would aid the elderly people, differently abled people and also the normal family. This developed intelligent robot companion named HomeBot (HB) that has ability to control the ambient conditions of the smart home environment based on the detected user behavior for improved the user experience. In order to enhance the interaction ability the interaction between the user and the HB is integrated with a vocal interaction module. After identifying the above mentioned current requirements which need to be addressed by an assistive home robot companion with the use of literature reviewing and brain storming sessions, HB was designed and developed which is capable of adapting the ambient conditions in accordance to the user behavior. The user behavior identification is facilitated by an artificial neural network that has been trained to detect the different postures of humans such as sitting and standing. Based on the identified user behavior, the robot controls the smart devices available in the home to realize the adaptation of ambient conditions. The smart devices and companion robot connect over a wireless network. Furthermore, voice interaction capabilities have also been incorporated to the robot companion to facilitate voice interaction based controlling of ambient conditions. A prototype of the system has been developed and the capabilities of the system have been validated experimentally. A robot companion capable of providing assistance and companionship to the users of all kinds such as to normal family, elderly people or to differently abled people will be the new technology to be experienced by the people in near future. Hence a robot companion who has the capability to provide both assistance, companionship and to provide smart home appliance controlling to provide most suitable ambient condition in the home in accordance to the user behavior was the aim to be achieved by this Research Project. Identification more postures and gestures by training the developed neural network and there by providing more user experience are proposed as further improvements.