Direct Adaptive Predictive Control for Wastewater Treatment Plant

Norazhar  Abu Bakar

Direct Adaptive Predictive Control for Wastewater Treatment Plant

Norazhar Abu Bakar

visibility

…

description

312 pages

link

1 file

Abstract

The purpose of this paper was to design a much simpler control method for a wastewater treatment plant. The work proposes a direct adaptive predictive control (DAMPC) also known as subspace predictive control (SPC) as a solution to the conventional one. The adaptive control structure is based on the linear model of the process and combined with numerical algorithm for subspace state space system identification (N4SID). This N4SID plays the role of the software sensor for on-line estimation of prediction matrices and control matrices of the bioprocess, joint together with model predictive control (MPC) in order to obtain the optimal control sequence. The performances of both estimation and control algorithms are illustrated by simulation results. Stability analysis is done to investigate the response of the system proposed when parameter changes exist. This project proves that subspace-adaptive method has a large number of important and useful advantages, primarily the application ab...

Figures (490)

Table | Condition of speech analysis and recognition. < http://www.raytron.co.jp/ >

performance among all methods including RASTA with DRA. Especially in the running-car noise environment at -10dB SNR, DRA improves the recognition rate with RSF by 31.15% while DRA improves that with RASTA by 20.75% only. Table 2 Recognition rates versus white noises for the estimation of feature extraction.

Table 3 Recognition rates versus running car noise for the estimation of feature extraction.

The first robot version needs a switch operation for the start point of speech recognition. Its operation results on low user ability. As our developed ASR v2 FPGA, the system had automatic speech detection under noisy circumstances. No one considers when speech recognition starts in this system. In this version, some answers by the robot and the controls by infrared light from the robot were implemented as robot reactions.

As the more cute and strong affinity for any persons, a baby doll ASR system was designed and developed. It was called “CHAPIT” which is shown in Fig.2. In 2008 International CES (Consumer Electronics Show) held in Las Vegas, the CHAPIT was introduced. Some of TV broadcasting introduced it in thier NEWS and his photos were then uploaded over 150,000 WEB site at that time. Fig.6 CHAPIT — ASR v2 FPGA system

liquid production. Here, Figure | depicts crude oil price scenario from 1970 to March 2012. It shows that the price has been fluctuated by many incidents around the world. The highest price was US$145.29/bbl in Jul 2008 during this interval.

Fig. 2 World marketed energy consumption; History and Projection, 1990-2035 [7]

Table 1 CO, emissions from fuel combustion from 1971 to 2009 (million tonnes of CO.)

Fig. 3 Non-OECD energy consumption, 1990-2035 The IEO 2011 report also unveils that energy use in non-OECD Asia (led by China and India) has the most robust growth of all the non-OECD regions, rising by 117% from 2008 to 2035 (Figure 3). However, strong growth in energy use also is projected for much of the rest of the non-OECD regions. With fast-paced growth in population and access to ample domestic resources, energy demand in the Middle East increases by 77% over the projection period. Energy consumption increases by 72% in Central and South America and by 67% in Africa. The slowest projected growth among non-OECD regions is for non-OECD Europe and Eurasia, which includes Russia and the other former Soviet Republics. Growth in energy use for the region totals 16% from 2008 to 2035, as its population declines and substantial gains in Energy Efficiency (EE) are achieved through the replacement of inefficient Soviet-era capital equipment [7].

emissions are likely to be increased 35.2 billion metric tons and 43.2 billion metric tons in 2020 and 2035 respectively. These emissions are mostly occurring in developing, non-OECD nations due to relying heavily on fossil fuels especially coal mine. IEO 2011 report also reveals that non-OECD countries’ emission is likely to be total 28.9 billion metric tons in 2035, or about 73% above the 2008 level. In contrast, OECD emissions are projected total 14.3 billion metric tons in 2035-only about 6% above the level in 2008. Figure 4 depicts the emission scenario and trends of OECD and non-OECD countries from 1990-2035 [7].

Fig. 5 Projected CO, emissions for four sectors in Malaysia

Fig. 8 Production, consumption and net export of Malaysia from 1990 to 2009; oil (a) and natural gas (b) [11]

Fig. 6 Asian pacific proven oil (a) and natural gas (b) reserve, Jan 2010

Fig. 7 Reserve of crude oil (billion barrels), natural gas (trillion feet’) and coal (billion short tons) [11]

Table 3 Percentages of coal in electricity generation [14] Proceeding of National Conference of Electric and Electronic Engineering 2012 Table 4 International crude oil and liquid fuels supply and consumption, Million bbl/d; 2007-2011 [20]

Fig. 9 Fuel consumption in Malaysian power sector from 2003-2008; Unit: KTOE [21] The system capacity was 3.15kWp and is connected to a 3-phase electricity system of the building. During the same year, two further grid connected PV systems were installed by BP Malaysia and Universiti Kebangsaan Malaysia (UKM) with the capacities are 8kWp 5.5kWp respectively (currently removed). The period 2001 to 2008 covers two Malaysia Plans (the 8" and 9" MP), which inaugurates three programs such as the Small and Renewable Energy Program (SREP) program, the Biogen Full Scale Model (Biogen FSM) Demonstration Project, and the Malaysian Building Integrated Photovoltaic (MBIPV) program. The aftermath of these initiatives from 2001 to 2008 (until September) are summarized below:

5.1 Ongoing Major Projects under TNB Table 5 RE based projects under CDM and its estimated reduction (tCO2e)

Table 6 FiT Rates for Solar PV, Small Hydro, Biomass and Biogas (21 years from FiT Commencement Date) [25]

Fig. 10 The second highest concrete-faced dam in the world, 2400MW hydro plant in Bakun, Sarawak; (a) - Eastward view of Crest from 2000 feet, (b) - Downstream view of Powerhouse [23] developed while the management has signed a joint venture agreement with Felda to set up a LOMW biomass plant in Jengka, Pahang.This RE plant should come on- stream by October 2012. TNB is also collaborating with Sime Darby Plantation to develop six potential biogas plants in Peninsular Malaysia with total capacity of 12MW, using palm oil mill effluent. Along with these, TNB Repair and Maintenance Sdn. Bhd. is engaged major contracts namely the Rehabilitation of Tenom Pangi hydro plant and the Operation & Maintenance (O&M) of Sandakan Power Station, 2400MW hydro plant in Bakun, Sarawak (Figure 10).

Fig. 11 Bridge illuminated with white GaN LEDs; could cut the proportion of UK electricity used for lights from 20% - 5% [28]

EE house requires smart design architecture and components. A whole-house systems approach can help successful architectural strategy incorporating EE in designing home. Some of the approaches are: advanced wall framing, roof technology, earth-sheltered home, passive solar home etc. Advanced house framing, also called optimum value engineering reduces the amount of lumber used and waste generated in the construction of a traditional wood-framed house. Home heating and cooling costs can be reduced through proper insulation and air sealing techniques. Earth-sheltered home provides a comfortable, tranquil, weather-resistant atmosphere. Especially it is designed to allow plenty of natural lights and organic architecture in curving walls and ceilings for temperature control. Figure 13a & Figure 13b depict EE building at PTM and Tsinghua University in Beijing.

Fig. 12 Solar heater fabricated with arrays of evacuated tubes Under the home EE appliances program, Australia is one of the pioneer countries to encourage for using EE appliances. The country has a survey report which states that around half of Australia’s eight million households use electric water heaters that produce up to three times the greenhouse gases of low emission technologies. The government is working to phase out for emissions savings of approximately 51.1 million tonnes of greenhouse gases over ten years under the electrical water heaters. Figure 12 shows the solar heater installed on the roof top.

Table-1: Markets that use flexography printing (Kleper, 2004)

Figure 1: Gravure printed images with jagged lines However, the flexographic printing process offers the possibility to print continuous fine solid lines (Yusof et al., 2006). While flexography is a simple process and the image carrier is cheaper compared with gravure, a vast number of parameters affect this process which will ultimately affect the quality of the printed images. However, current research on printing plate technology (Yusof et al., 2005) and investigation into critical parameters such as anilox rollers, ink rheology etc (Yusof et al.,2007) help to comprehend the impact and the affects of these and other parameters on the final printed images. Further research on the application of finite element analysis (FEA) including both linear and non-linear models provides complementary work to predict the printing plate deformation (Yusof et al., 2007) and in understanding the ink spreading mechanism (Yusof et al., 2007).

Figure 2: Schematic of Flexography printing process Nevertheless researches in the flexographic printing are continuously carried out where the process itself is constantly being developed and now have a capability that approaches those of the high-speed processes. It is used dominantly in packaging printing and mainly used for graphic purposes but has recently evolved to many other technique depending on the applications as seen in Table-1. And since it is widely used as packaging printer, it has a big advantage in printing RFID antenna and other applications onto the packaging label itself.

Figure 3: RFID tags components Readers - An electronic device that uses radio frequency to read or interrogate the tags. When an RFID reader and tag communicate, two things happen; they share energy in the form of radio frequency which allows the exchange of information. 2.3 Finite Element Analysis On Multiple Solid Lines

Figure 2 : Sample image of Cryptosporidium cyst

Table 1: Invariant feature vectors for coconut images

Table 2: Invariant feature vectors for fronds images From Table 1, the invariants for the @1 are ranges from 0.19 until 0.22. Ranges for @2 are insignificant. Hence, the identification of the coconut object only used the @1 invariant values but some of the image used has a substantial effect on the feature values. Error may occur in detecting the coconut images when the invariant of the fronds images are quite similar to the coconut. Meanwhile, some of the coconuts images are failed to detect as the coconut because of the invariant are out of the range. This problem occurs because the coconut image is too small and consist a lot of fronds.

Je SAVER BREN EPNVEINYAANSG 2 Image analysis has already proved to be a potentially alternative tool to overcome the drawbacks associated to the micro-organisms visual identification and quantification. (Y.P. Ginoris et al, 2007). The water cyst detection using Hu invariant moment consists of three phase to implement the project. MATLAB R2008a is used to create the system. The first phase is project research. In this phase, some study on types of bacteria found in water is done. There are two types of water cyst used in this project which is Giardia and Cryptosporidium. There is some difference in shape between these two types of bacteria. The Cryptosporidium cyst has a circle shape while the Giardia type has an oval-like shape. For the second phase is system development. The flow chart shown in figure 3 was started with getting the image. Cyst image sample was getting from the external sources. Stained samples are viewed with an epifluorescence microscope as shown in figure 4. mage processing is the sequence of image starting from the conversion of RGB image to grayscale image form as shown in figure 5. Then, it will filter using a median filter that is often used in image processing to reduce the noise as shown in figure 6. Finally, the filtered image will be converting to binary image as shown in figure 7. The next is feature extraction process. Hu invariant moments will produce the seven invariants which is will construct the central moment. nvariant moment is a technique that introduced by Hu. (Hongbo Mu, 2009). This technique is very useful in application to describe the region shape. Neural Network techniques are used to classify the types of cyst either Giardia or Cryptosporidium type. Currently, the uses of Neural Network technique have been use in pattern recognition. Pattern recognition has been widely used as classical statistical methodologies and applications of Neural Networks have been applied since several decades. The classification process involves two set of water cysts data which are admin train and user test data. Both of these data must be created and save in excel file in order to complete the classification process. The function of admin train is to train the Neural Network. It consists of all the parameter needed in order to classify the cyst type. User test is to record all the cysts parameter during running the system. For third phase is analyzing the result.

Table 3 : Invariant feature vectors for Cyst images The sample image of Cryptosporidium and Giardia cyst were used to get the range of the invariant. In this part, it covers the data that obtained from the project has been conducted. Hu Invariant moment technique will produce seven moments that will be recorded and use for grading process. The parameter such as area, perimeter and centroid also used as input for Neural Network to classify the types of cyst. The parameter are shown in Table 3. Cryptosporidium type was classified with number ‘1’ and Giardia types are classified with ‘0’.

Figure 8: Graph of invariant 1 (M1) From the table 3, the invariant for M1 are in some range 1.2581 to 2.0928 for Giardia cyst and 4.1041 to 7.422 for Cryptosporidium cyst. The invariant for M3 are in range 1.5103 to 23.362 for Giardia and 54.0287 to 527.1612 for Cryptosporidium. For M4 are in range 1.8513 to 10.4955 for Giardia and 77.4198 to 561.3904 for Cryptosporidium. The area range for Giardia is in 2754 to 4340 and for Cryptosporidium are in ranges 606 to 1171.

Figure 10: Graph of invariant 4 (M4) Figure 9: Graph of invariant 3 (M3)

B. Conclusion Figure 13 : Guide User Interface (GUI) In this project, the invariant value for M1, M3, M4 and area are used in detecting the types of cyst. Both types of bacteria have significant differences. Figure 8 until figure 11 shows the graph range for invariant and area that was used for classification process either Giardia or Cryptosporidium. For M2, M5, M6, M7, perimeter and centroid onwards the invariant are insignificant. This insignificant value still used for training process. Figure 12 shows the example graph for insignificant result of perimeter. Based on the graphs, the range of parameters for Giardia is from 353 to 568.5, while for Cryptosporidium is the range from 445.4 to 472.5. The graph shape shows the values of parameters are not stable. So the perimeter will not be used for classify the cyst.

Figure 1: Basic architecture of MLP neural networks

of the net input from negative to positive infinity. Figure | shows the basic MLP neuron network structure with R as the number of inputs. Each input will be adjusted by the weights, w. Total input weights, w and bias, b would lead to a transfer function, f. Bias, b is approximately the same as the weights, w but the bias has a value | as constant value. Weight value, w and bias, b is a scale parameter which can be modified. Usually, the MLP network will use log-sigmoid transfer function as shown in Figure 2 in the hidden layer. In most neural network applications, log-sigmoid transfer function is often used in object recognition. Transfer function will then calculate the value of output from the net input. This operation of log- sigmoid transfer function (log-sig) will produce an output between 0 and | which indicates the value of the net input from negative to positive infinity.

In this project, the distribution of data can be divided into ;

Value obtains from the Oscilloscope: Based on Figure 3.1.1 and Figure 3.1.2, it could be seen that the shape for Triangular wave is same. As the signal being validate using the Oscilloscope, the actual value of the frequency drop a little bit from the given frequency which is 50Hz to 49.80Hz. Meaning that;

Figure 3.1.1: Triangle wave shown from Matlab Figure 3.1.2: Triangle wave from the Oscilloscope

Figure 3.1.1 shows the output result of triangle wave from the Matlab window with the input frequency of 50Hz and amplitude of 1 and -1. On behalf, Figure 3.1.2 shows the validation result of triangle wave as viewed from an Oscilloscope.

Figure 3.2.1: Sinusoidal wave shown from Matlab Figure 3.2.2: Sinusoidal Wave shown from the Oscilloscope

Figure 3.3.1: Triangular wave shown from Matlab Figure 3.3.2: Rectangular wave shown from the Oscilloscope

Figure 3.3.1 shows the output result of rectangular wave from the Matlab window with the input frequency of 150Hz and amplitude of | and -1. Then, Figure 3.3.2 shows the validation result of rectangular wave as viewed from an Oscilloscope.

Value read from the Oscilloscope: Based on Figure 3.3.1 and Figure 3.3.2, it could be seen that the shape for Rectangular wave is same.

640,538 GWh (79.9%) of the electricity produced, with coal and natural gas dominating. Renewable sources — large and small hydro-electric power units, wind energy and biomass power — contributed 142,654 GWh (17.8%) to the total. The largest contribution (13.6%) came from large hydro-electric power units. Nuclear energy contributed only 18,636 GWh (2.3%).

Next, we predict India’s future population. Table 2 presents data on the population at 10-year intervals, starting in 1951. The percentage growth over the intervals is also given. Data for 1951-2001 are actual values, while those for 2011 are preliminary values based on the on- going census’. It is seen that the population has grown from 361.1 million in 1951 to 1210 million in 2011. from 361.1 million in 1951 to 1210 million in 2011.

Table 3. Electricity in India in 2009 As per the surveys conducted by International Energy Agency’ the following data regarding the electricity supply of India for the year 2009 is tabulated below.

Figure 1. Causal loop diagram for population changes population on birth rate and death rate. A causal loop diagram is a simple map of a system with all its constituent components and their interactions. By capturing interactions and consequently the feedback loops (see figure below), a causal loop diagram reveals the structure of a system. By understanding the structure of a system, it becomes possible to ascertain a system’s behavior over a certain time period. The following shows an example of the causal loop diagram used in the modeling describing the dependency of the variable population on birth rate and death rate.

Figure 4. Stock and flow diagram to study the contribution of various power supply source to meet the electricity demand

The following is shows the stock and flow model of e causal diagram shown in Figure 1. Figure 3. Stock and Flow diagram for the population changes

Figure 6. Forecast of utilization of non-renewable sources in future variation of population in the same time period.

Figure 5. A forecast of the population of India till 2025

We can observe a great gap between the demand and further increase in the future if we rely only on non-

Figure 7. Comparison between the supply and the demand for electricity and their difference Figure 7. Comparison between the supply and the figure are tabulated below in Table 4. The figure 7 shows the comparison of the electricity supply, demand and their difference over a period of 16 years from 2009 to 2025. The observations from the figure are tabulated below in Table 4. The figure 7 shows the comparison of the electricity

Fig 1. Cell culture in the absence (left) and presence (right) of liquid crystals on gold substrate.

Where ro;, and rojp are the reflection coefficients for s-polarised and p-polarised light, respectively. 8, and 0, are the angle of incident and refraction as obtained from Snell’s law [12].

Fig 3: An incident light hits interface of medium 0/1 at an angle 8, and reflected in medium 0 at angle @,. Consider a ray of incident light hits at the interface between medium 0 and 1. Each has index of refraction of No and N, respectively. The incident wave vector with an amplitude of E, arrived to the interface at an angle of 0 with respect to the normal plane (Figure 3). The Fresnel’s coefficients of the reflectance of the s- polarised and p-polarised light is given by

In high contrast, the adherent cell was found less spread and clearly defined narrow concentric rings were seen for cells attached to the liquid crystal (Figure 2B). This gave an indication that the the stress of the focal adhesion was around edges (narrower band) of the cells membrane when cells were cultured on the liquid crystals. In fact, this narrower stress was formed by a more uniform arrangement of smaller focal adhesions as evidenced by the vinculin staining (Figure 2D).

Fig 6 The reflectance coefficients of the s-polarised and p-polarised light incident at the (a) glass/gold/air and (b) glass/gold/LC systems.

The scaling and wavelet functions are calculated by taking the inner product of the coefficients and data input values. In the last iteration, data input of s[N] and s[N+1] are not exist (they are beyond the end of the array) and cause the edge problem. To handle this edge problem, the data set is treated as it is periodic. The process of Daubechies wavelet using both scaling and wavelet functions are given as follows. Qtan 1-

Table 1. Resources utilisation and overall proposed architectures performance on EP2C15AF484C6 for N = 4.

Figure 3. Proposed system architecture 1-D Daub4 flow diagram with N inputs sample for direct mapped architecture.

Figure 4. Comparison of chip floor plan for N = 4.

Fig | Layers of the Earth’s atmosphere from the surface of the Earth [1] The Earth’s atmosphere is a thin layer of gases that surrounds the Earth. It composed of 78% of nitrogen, 21% of oxygen and 1% of other gases such as argon and carbon dioxide. This thin gaseous layer insulates the Earth from extreme temperatures; keeps heat inside the atmosphere and also blocks the Earth from the sun’s incoming ultraviolet radiation. The ionosphere, which is found from the altitude of about 60km to about 700km 1], contains up to four layers of free electrons which can enable long-distance radio communication. The Earth’s atmosphere consists of a few layers as can be seen in Figure | below.

Fig 2 Cycle 24 Sunspot Number Prediction [7]

Further, the analysis presented here includes worst case scenario to evaluate system performance. The simulation carried out is extension of the two cell model to represent MC-CDMA in time duplex mode [3]. The MC-CDMA air interface allows high-capacity networks and robustness in the case of frequency- selective channels, taking benefits from CDMA capability offered by the spread spectrum technique, and MC modulation as orthogonal frequency division multiplex (OFDM). A_ possible generic downlink transmission scheme is depicted in Fig.2. Each user data can be simultaneously processed at the spreading step before MC modulation. In the following, due to their good properties for the downlink, Walsh-Hadamard (WH) spreading sequences will be considered. The presented MC-CDMA configuration is based on the transmission of multiple data per MC-CDMA symbol for each user. Data d’; (n) denotes the ith, 1 <i < N,, data transmitted by user j, 1 <j < Ny, in the nth MC-CDMA symbol[4].

Fig.3 Transmitter schematic of MC-DS-CDMA In order to transmit high-rate data with sufficient processing gain, the chip rate of the DSCDMA system should be significantly higher than the practical limit imposed mainly by the processing speed and power consumption of state-of-the-art electronics. In this case, parallel transmissions of DS-CDMA signals using the OFDM structure can be a solution[5].

Fig.2, MC-CDMA Transmitter and Receiver Structure Obviously, MC-CDMA system offers high flexibility in resources (spectral efficiency, number of users) allocation which consequently induces large design spaces. As a result, high-level design methods are convenient in order to deal with such complexity and for efficient implementation [4].

Fig.3 shows the transmitter structure of a MC-DS-CDMA system. The N consecutive input bits of the jth user, b’;, i =0- --N~1, are serial-to-parallel converted first. Then, each bit b’; is spread by the jth user’s spreading sequence in the time domain. The other operations are identical to those of MC-CDMA. In other words, in the MC-CDMA the data bits arriving at a rate of 1/T are first spread and hence have a rate of N/T, before reducing the rate again to 1/T . By contrast, in the MC-DS-CDMaA the bit rate of R, is first reduced to R,/N and then we produce the rate R,G/N using spreading [5]. Fig. 4(a) shows the typical spectrum of the MC-DS- CDMA signal operated using the schematic of Fig.3. The subcarrier separation, fg, meets the orthogonality condition of:

of TD-SCDMA was standardized first in CCSA around the Release 5 timeframe of 3GPP, and included into Release 7 in 3GPP. The approximate relationship between the versions of TD-SCDMA released by 3GPP and CCSA is shown in Fig.5 [7]. TD-SCDMA has several different features from UMTS FDD, including a narrower bandwidth and hence lower data rate, the possibility to support asymmetric downlink/uplink (DL/UL) allocation of transmission resources, and, being a TDD system, the potential to benefit from transmit/receive channel reciprocity. Therefore, several TD-SCDMA-specific technologies have been developed to exploit these features [7]. maa. eee oe

Fig.6 A typical smart antenna architecture.

Table 1. Main characteristics of different MC-SS concepts

Fig.7 General principle of MC-CDMA and MC-DS-CDMA systems indoor, and fixed and broadband wireless access (F/BWA). In addition to system-level analysis, a multitude of research activities have been addressed to develop appropriate strategies for detection, interference cancellation, channel coding, modulation, synchronization (especially uplink), and low cost implementation design[8].

Table 1 : Comparison the Performance of Three PBG Structure [2]

Figure 1 : A fabricated PBG Microstrip sample and size has been proposed to analyse the performance of these materials such as periodic pattern of circular, rectangular pattern, square pattern and cross pattern. For examples, from research [2], they make an analysis with three difference periodic geometry of pattern. The geometry pattern involved are square, cross slots and a scalariform geometries. The three structures based on the same substrate and same period. A sample with a pattern arrangement of 7 x 7 elements is used as shown at Figure 1 and the result shows at Table 1. It shows the performance of the PBG structure incorporated with microstrip line with three different geometries and the scalariform cross slot geometries has a higher bandwidth.

Figure 2: A view of td-test.geo file in XY with absence of PBG material on the ground plane. Figure 3 : A snapshot on the windows shows the input (probe 1), output(probe 2 ) and S>;

Without Any Material Of Pbg At The Ground Plane

Figure 6 : A snapshot shows 3mm square PBG placed at the ground plane Figure 5 : A snapshot shows 7 x 7 elements of 2 mm PBG on the ground Plane

Figure 4 : A snapshot shows 7 x 7 elements of 0.9mm square PBG at the ground Plane

Figure 7 : A snapshot shows the output of S., with different size of PBG.

Figure 9: A snapshot shows 3mm Scalariform PBG place at the ground plan

Figure 8: A snapshot shows 3mm Square PBG placed at the ground plane The simulation had been made with different shape but same size of PBG material. Based on the simulatior earlier, 3mm square PBG provides us with bette: performances in bandwidth and depth of stop-banc compared to 2mm square PBG and 0.9 mm square PBG Therefore, 3mm of different shape of PBG will be investigate to study the effect of shape in influence the bandwidth and depth of stop-band. The shape involves ir the simulations are square, scalariform and ‘stage.

Figure 10: A snapshot shows 3mm ‘Stage’ PBG pattern at the ground plane Proceeding of National Conference of Electric and Electronic Engineering 2012

Figure 11 : A snapshot shows the output of S21 with different Shape of PBG.

Figure 13 : A snapshot shows 3mm Square PBG pattern at the ground plane with non periodic distance between PBG

ure 12: A snapshot shows 3mm Square PBG pattern at the ground plane with periodic distance between PBG

Figure 14 : A snapshot shows the output of S21 with periodic & non periodic distance of PBG.

Figure 1: Magnetometer and Backpack Assembly The magnetometer is very simple to assemble. As referred to Figure 1, the sensor is installed on top of the staff. The sensor is connected to the console through sensor cable. The following picture shows the configuration of magnetometer in walking mode.

number of satellites currently in view with respective satellite ID. Figure 2 shows the Garmin GPS navigator attached to the GSM-19GW console. Figure 2: Garmin GPS Navigator Attached to the GSM-19GW Console

Figure 5: Open Field Survey Area Figure 3: WARAS survey area

Figure 6: UTHM Library Survey Area Figure 4: Tun Fatimah College Survey Area

Table 1: Space Weather Alert Description

Table 2: IGRF-11 magnetic field value at different places in UTHM

Source : (Space Weather Prediction Center,2005)

Figure 8 illustrates the Earth’s magnetic field variation at different places in UTHM. It can be seen that the magnetic field value at UTHM Library is radically high compared to Tun Fatimah College, WARAS and Open Field respectively. The reason is due to high cultural interference contribution during magnetic field survey conducted such as building and cars. TO a. rr ee ee oo survey location .Therefore, it can be concluded that the cultural interference influenced Earth’s magnetic field magnitude. Table 3: Earth’s Magnetic Field Variation at different places in UTHM

4.4 Comparison Of Actual Magnetic Field Data And Igrf-11 Magnetic Field Data At Different Places In Uthm

Figure 9 shows the difference value between actual magnetic field data and IGRF- 11 magnetic field data at certain places in UTHM. The difference value is obtained by subtracting IGRF-11 data from actual magnetic field data. As predicted, UTHM Library has the highest difference value of magnetic field. The sequence followed by Tun Fatimah College, WARAS and lastly Open Field.

Figure 10: Contour map at UTHM during morning (top left), noon (top right) and evening (below) on 1‘ November 2011

Figure 11: Contour map at UTHM during morning (top left), noon (top right) and evening (below) on 12"™December 2011

Figure 1 Configuration for antenna calibration and test site evaluation This paper presents antenna calibration using SSM and RAM in a 3m OFTS and the evaluation that OFTS as well as 3m SAC. The configuration for antenna calibration and site attenuation measurement for test site evaluation was the same as shown in Figure 1.

where SAj7, SA;3, and SA>3 are site attenuations for antenna | and 2, 1 and 3, and 2 and 3 pairs respectively, while Epmax for each frequency are available in [1]. The results are given in Table 3. Antenna factors for the three antennas were calculated using the following equations:

Antenna factors of the antenna under calibration were calculated using the equation below: Table 4 RAM data

Table 5 Antenna factors obtained using RAM The antenna factors obtained using both methods were compared to the antenna factors provided by the manufactures as listed in Table 6.

Table 6 Antenna factors from manufacturers

Table 10 SAC evaluation data. NSA for OFTS and SAC were calculated using equation 2 and compared to theoretical NSA available in [2] as given in Table 11.

Table 11 Calculated OFTS NSA, SAC NSA and theoretical NSA.

Figure 2 Comparison among NSA of SAC, OFTS anc theoretical.

Table 2 Bandwidth requirement and power requirement of OOK, L-PAM, L-PPM, DPPM and DAPPM. The normalized optical power versus bandwidth required for OOK, PAM, PPM, DPPM, and DAPPM is shown in Fig. 3. Each point for PAM, PPM, DAPPM(A=2), and DAPPM(A=4) represents the level L={2,4,8,16,32}.

Table 1 Symbol mapping of OOK, PPM, DPPM, DAPPM (A=2, L=4), and DAPPM (A=4, L=2) for resolution of M=3.

considerations in the order of importance from optical wireless communication stand point. Fig. | Pulse modulation tree [7].

A set of DAPPM waveforms is shown in Fig. 2. The average number of empty slots preceding the pulse can be lowered by increasing the number of amplitude levels A thereby increasing the achievable throughput in the process. When compared with similar modulation techniques, a well designed DAPPM will require the least bandwidth. DAPPM suffers from a high average power and a large DC component, thus restricting its use to applications where power is not a premium. It is also susceptible to the baseline wand.

Fig. 3 The normalized optical power and bandwidth required for OOK, PAM, PPM, DPPM and DAPPM. 3. Results and Discussion

This system has two inputs (BER and Rate), and one output (Levels). After finish setting the fuzzy sets, membership function and rules, the rule viewer for System A is shown in Fig. 5. The five steps for fuzzy inference process are setting the Fuzzify inputs, applying the fuzzy operation, applying implication method, applying the aggregation method, and perform the centroid defuzzification. Finally, the output value will be calculated by Fuzzy Inference System (FIS) after completing the whole process. In this Figure, BER = 2 and Rate = 0.5 is chosen as sample inputs. The required level change is 2.69.

Fig. 4 Block diagram for BER degradation and variation rate to modulation level (System A). and rate variation to modulation level and System B is the BER level and variation rate to modulation state

The rules for System B can be expressed as follows: The BER is not only be affected by noise, transmitted signal power, and modulation level but is also affected by modulation state. An fuzzy inference system was set up to solve this problem and named System B. BER level and rate are set as the input fuzzy set, while the Amplitude level (A) and the differential pulse position change will set value (L) as the outputs fuzzy sets as shown in Fig. 7.

Fig. 6 A three-dimension curve of surface viewer for system A.

Fig. 8 Rule viewer for fuzzy controller to change the system status and stabilize the BER.

Fig. 9 The surface of amplitude level (A) versus the BER level and change rate.

Fig. 10 The differential pulse position (L) versus the BER level and change rate.

In the recent years, many algorithm and protocols about energy-efficient have been proposed. The cluster- based model is better than single-hop or multi-hop model. The cluster-based routing algorithm was proposed by [4] that optimizes the energy efficiency in WSNs. This algorithm known as Low-Energy Adaptive Clustering Hierarchy (LEACH) where the idea of this algorithm is the cluster members elect cluster head to avoid excessive energy consumption [8]. LEACH consists of 2 phases: set-up phase and steady-phase as shows in Fig. 2. In the set-up phase, sensors may elect randomly among themselves a local cluster head with a certain probability. By doing so, the network may balance energy dissipation across the whole network. The optimal number of cluster heads is 5% of the total. After the cluster heads are selected, the

The cluster heads are selected sequentially depending on the sorted list. The list is updated for neighbours based on already selected cluster heads. The process continues until all the nodes have a cluster head. The “cluster-head” assignment to the rest of the nodes is done in a sequential fashion from the top of the stack of cluster heads selected. Each node is then checked against the selected cluster head list to find the nearest cluster head. If the nearest cluster head is in its transmission range then the node is assigned to that cluster for the current round. Round proceeds as in LEACH.

Fig. 4: The Number of Rounds when First Node Dead with Different Number of Nodes From Fig. 5, the Half Node Dead (HND) occurs for LEACH and Modified LEACH is not giving a lot of change. Although the number of nodes increases, the difference between LEACH and Modified LEACH does not differ greatly.

Fig. 5: The Number of Rounds when Half Node Dead with Different Number of Nodes

CLALMOLTIIOOLUETE 11 A ULI? Lit UI Migital RILVILUILIIIVIIL, such as telecommunications, multimedia and integrated services digital networks, and video recordings and storage of medical images and archives fingerprints. The number of pieces that require the representation is encrypted must be less than that necessitated the original image. So that it can use less time communication or storage, and also to maintain the space in the memory unit and to facilitate the process of transferring the image from one place to another over the Internet. In the ideal image compression system includes three components as directly linked as shown in Fig 1. [4,5]. Fig 1 Signal encryption / perfect picture without losing

Fig. 3 Three images with gray level by the use of compression techniques Then convert the colored image to the gray level as shown in Fig 3, where the image is of type unit8 and so that we can deal with them mathematically to determine the degree of gray image is converted to type double.

Fig.2 Three color images that were used in the program We will ask for different pictures (A, B, C) of the file stored by each image to create a database of images, where the image A is (a clown) and image B (image scraps of paper) and C is (image nature), shown in the Fig om

Fig. 5 Represent the image resulting from the application of DCT on the original image by changingthe size of the area

Fig. 6 Represent the image resulting from the application of DCT on the original image by changingthe size of the Table 2 shows the value of PSNR and bpp to image paper scraps by using DCT

Table 4 shows the value of PSNR and bpp to Clown image by using FFT

Fig. 7 Represent the image resulting from the application of FFT on the original image by changingthe size of the area Table 3 shows the value of PSNR and bpp to image Nature by using DCT

Fig. 8 Represent the image resulting from the application of FFT on the original image by changingthe size of the area Table 5 shows the value of PSNR and bpp to image paper scraps by using FFT

Table 6 shows the value of PSNR and bpp to image Nature by using FFT Fig. 9 shows the relationship between DCT and FFT by Zonal Sampling Fig. 8 Represent the image resulting from the application of FFT on the original image by changingthe size of the area

End to End Transmission Delay (second) Figure 2: Packet Delivery Ratio

Figure 4: Routing overhead for various speed of node mobility

Fig. 1 Auroral oval on 24 May 2008. Location of Alert | superposed on the map. (Source: http://sd- www.jhuapl.edu) (MLT) midnight is ~2216 UT. Fig. | shows the location of Alert superposed on the map illustrating the auroral oval location. The map is generated for 24 May 2008 at 1340 UT during the period of low geomagnetic activity. Alert is located far from the auroral oval and very near to the geomagnetic pole, making it an interesting place to study the ionospheric variability.

Fig. 2: Plot of elevation angle versus $4 for all data > 25° divergence versus S4 showed that most of the data were above the dividing line, which is referring to potential multipath cases and not the scintillation ones.

Fig. 3: Plot of the standard deviation of carrier/code divergence versus S4 for all data > 25°

were between 0.2 and 0.3. Therefore, all S4 were considered as insignificant. This is in agreement with the work done in [7] and [5] where observations covering the periods from 2007 to 2009 at high latitude stations discovered that amplitude scintillations were very low during those periods.

Fig. 1 Schematic diagram of regeneration based on saturation of FWM. In this paper, we experimentally study noise suppression and addition in the regeneration based on saturation of FWM for OOK and DPSK signals. We investigate the regeneration performance when using continuous wave (CW) pump, as in [10], and pulsed pump. Signal performance is investigated by measuring BER versus threshold voltage. Progressive development in optical transmission technologies allows the growth of transmission speed and capacity. As the optical transmission system moves toward higher bit rates and more advanced technologies evolve, signal degradation becomes more dominant. Regeneration can refine the degraded signal and improve transmission performance. Conventionally, electrical repeaters are used for optical signal regenerations, which need optical-to-electrical and electrical-to-optical signal conversions. All-optical regenerator is an alternative to electrical repeater, which can work ultrafast and less complicated.

Fig. 2 Power transfer function of unmodulated signal pulses.

Fig. 4 BER evolutions vs threshold voltage of DPSK signal when input OSNR is 20 dB/0.1nm.

Fig. 1 Signal waveforms of the nLVSD driver. This paper will start off with the introduction of the mLVSD driver as well as the interconnect model and associated receiver used for this work. This is followed by the comparison analyses and discussion between the diode-connected driver schemes, namely, the nLVSD, mLVSD, MJ and DDC driver schemes. It is to be acknowledged that the DDC driver is also a diode- connected driver which was analysed in [5] and compared with the MJ driver in terms of performance and power consumption in [6]. The design parameters that are to be compared are delay, power consumption and noise immunity.

Similar to the nLVSD driver, the mLVSD employs a diode-connected transistor pairs at the output. However

The new low-swing driver proposed in this paper is known as the mLVSD driver. This driver is an improved version of the nLVSD as it incorporates all of the essential components of diode-connected driver as in the nLVSD circuit with similar structure as shown in Fig. 2 and Fig. 3. Even though both drivers incorporate leakage control mechanism, but the nLVSD driver implements leakage control transistor whilst a pair of pass transistor is encapsulated in the mLVSD driver circuit. Both of these mechanisms not only contribute to the reduction of the leakage current but also in the size of the output voltage swing.

The low-power high-speed level restorer used for this work is a device which resembles a Schmitt Trigger, as Fig. 5 The architecture of low-swing signalling scheme.

Fig. 4 The Schmitt Trigger receiver configuration. shown in Fig. 4. This level converter has the same characteristics as the Schmitt Trigger and provides a fast signal detection speed compared to a standard inverter. The Schmitt Trigger receiver, also known as the ScTr level converter, can turn a slowly varying input signal into a clear digital output signal. It provides a significant improvement in terms of delay and power consumption due to its fast transitions at the output by suppressing the direct-path currents. Even though it consumes slightly more area than a simple inverter, better power and delay performance is readily traded off against a small increase in area. The Schmitt Trigger receiver is followed by a simple inverter chain to provide gain for the scheme. The inverter chain is minimally sized. The receiver inverters are implemented with a channel width Wn = 0.27um and Wp = 0.8lum.

Fig. 6 The comparison of delay between the low-swing signalling schemes for different range of interconnect length. The performance of the low-swing signalling schemes were analysed and compared with respect to the main design parameters, namely delay, power consumption and signal-to-noise ratio. The first design parameter considered was the overall delay of the driver schemes. The analysis was performed by increasing the length of the interconnect, from 1 to 10mm, between driver and receiver and determining the value of signal delay. A plot of the delay against interconnect length is shown in Fig. 6.

increase in power consumption compared to the nLVSD driver scheme, and 7% improvement compared to the MJ and DDC driver schemes. At 10mm line length, the nLVSD driver scheme produces 43% improvement in power consumption compared to the MJ and DDC driver schemes. Fig. 7 The comparison of dynamic power consumption between the low-swing signalling schemes for different range of interconnect length.

Fig. 2 XRD analysis Based on the XRD diagrams above display the zinc give has the highest peak which clearly is the dominant elements. But there are also peak is high apart from the zinc because the zinc oxide also displayed relatively high peak. This is because zinc has been oxidized during sintering process. However this is an initial preparation step direction of zinc as control experiment.

Fig.1: A basic electrical stimulation system paralyzed muscles to make contraction. The main objective of FES in injuries to the central nervous system is the substitution of the absent bioelectric activity with an appropriately formed series of electric pulses, generated by a stimulator, or the elimination of the hyperactivity in paralysis and spastic paresis (Pasniczek and Kiwerski, 2004). Basically, two electrodes are essential to close the current circuit of the stimulation system as shown in Fig.1.

Fig.2: Typical circuit of a transformer-based FES (Cheng et al., 2004).

Fig.3: Switching resonant circuit FES (Cheng et al., 2004).

Fig.5: The illustration of proposed stimulator system Proceeding of National Conference of Electric and Electronic Engineering 2012

Figure 2.Recycling Folded Cascode OTA [1] Fig.2 shows recycling folded cascade structure which is proposed in [1]. The input pair is split into half ( M3A, M3B, M4A, M4B) while ratio of current mirror

Fig.5 shows the frequency responses of both RFC and proposed OTA with 5 pF load capacitor respectively. The circuit has been simulated using LTSpice and BSIM MOS transistor models (level 54) with well 50nm CMOS process. Both folded recycle OTA and proposed OTA are designed to have same 100uA tail current and power supply | V. Fig.4 shows that the DC gain is improved by approximately 5dB, however the phase margin of the proposed architecture is slightly lower compared to RFC. This is due to the increased capacitance at the cascade nodes, increasing pole and zero. The dc gain of the new OTA is significantly increased from 52 to 59 dB. The unity gain frequency of the RFC is 260 MHz with 48° phase margin, while new OTA has 230 MHz with 40° phase margin. The effect of non dominant poles decrease the phase margin of the new OTA, but it still a sufficient phase margin is achieved. Step response of the new OTA is shown in Fig.4 to confirm the stable operation of the circuit. Simulation results show there is a slight overshoot due to 40° of phase margin.

The large signal transfer characteristic of the OTA for unloaded output is shown in Fig.6 and the voltage swing is relatively wide.

The simulation results is shown in Table 1 comparing the performance of recycle folded OTA with proposed architecture.

Figure 1: System overview Figure 1 show the configuration of the project respectively.

The ultrasonic transducers are extremely sensitive and have specific mounting requirements [3 4]. To protect the sensors from foreign matter they have been enclosed in a solid plastic container. Holes have been drilled in the container to allow pressure equalization [5].

Human body joint angle usually measured by using goniometer and lately new technique using gyroscope developed in term of better accuracy[5]. Bakshi, S. et al. (2011) compared between various techniques that previously used in human body joint angle movement measurement described in table 1 before it come out to use IMU unit which consists gyroscope and accelerometer[1].

Joint angle measured by comparing the different angle between gyroscope 2 at shank and gyroscope | at thigh. Gyroscope will produce 3 axes which z-axis as a reference axis. At the origin, angle from z-axis to x-axis and y-axis will be at 90°. While SCI patients move their leg, the angle from reference z-axis to new x-axis with y-axis as a pivot will change. This change will acknowledge as movement angle for each thigh and shank. The absolute angle measurement for each MEMS vibratory gyroscope that proposed by Friedland and Hutton in 1978 been proved by Piyabongkarn, et al. (2005) in[6]. This method is used in this project to generate angle for gyroscope 1 and gyroscope 2. SCI patients straight up their leg like in figure 2 that make orientation of gyroscope | and gyroscope 2 same and there no different in angle which mean patient do not bend their knee and there is 0° of knee bend.

Figure 3: Gyroscope Angle at Knee Bend where # is angle of knee joint, aw is angle from Z, to X, at gyroscope | at thigh and # is angle from Z, to %, at gyroscope 2. This @ and # angle get from theorem and equation in[6].

Fig. | Example mammograms, where the breast density increases from (a) BIRADS I to (d) BIRADS IV. Fig. | shows example mammograms of each class (the mammograms are extracted from the MIAS database). Note how the internal density of the breasts increases from BIRADS I (left) to BIRADS IV (right). It should be noted that besides density these BIRADS classes also included patterns that can be described as various textures. As such, it seems appropriate to include both aspects in an automatic classification approach.

Fig. 3 Breast cancer mammogram clustering. Segmentation using clustering method is the easier way for developing image processing. Clustering can be considered the most important unsupervised learning segmentation, which deals with finding a structure in a collection of unlabeled and spread plot image. Clustering simply define as the process of organizing objects into groups whose members are similar in some way. A cluster is therefore a collection of objects which are similar between them and are dissimilar to the objects belonging to other clusters. This example is shown in Fig. 3 where applied for breast mammogram.

Fig. 4 Quadtree decomposition, all points of a square have the same homogeneous brightness value. The numbers of the field are simultaneously at the depth of the division. *. To encode each must be _ created’ by subdividing land reexamined for homogeneity. This means a considerable amount of time when decomposition as shown in Fig.4. In contrast, the decoding of the image can be built quickly and directly.

Then homogeneous brightness region based enhancement model previous method where neighboring pixels group into one areas that have the same color or shade (gray scale image) by attribute of visual perception in which a source appears to be radiating or reflecting light. Fig. 5 show region of cancer pixel area used in homogeny brightness region. Eliminate muscle and cancer regions the process continues by employing Fuzzy-C Mean clustering. Fig. 5 Region of cancer pixel area used in homogeny density.

glandular. However, the genus class of abnormality clearly shows the Spiculated masses have the higher processing time for the rest of classes.

Fig. 7(a) is the original Mammogram image and 6.(b) is the Segmented image. The image is Cluster into five dominant regions as shown on 7.(c) - 7.(g).

Fig. 8 Result processing cost for different classes.

The PL spectrum of the deposited ZnO films were analysed based on PL spectra measured in room temperature. The ZnO film exhibits one recognizable UV peak at 380 nm as shown in fig. 2. The UV emission peak at 380 nm which is related to the near band edge emission is due to the excitonic recombination in ZnO [17]. Fig. 2 Room temperature PL Spectra of the deposited ZnO film.

Fig. 3 I-V characteristics of the ZnO based UV photodetector.

Fig.1.Electromagnetic field on waveguide surface known as “evanescent field”. Photonics technologies as silicon-based photonics have vastly improved in the recent year mostly in optical communication and optical interconnects. [1]. Integrated photonics devices offer an important solution in improving the limitation of speed, packaging, power dissipation and sizes in microelectronic circuits/devices. [2]. One of the fundamental photonics components is waveguide and_ silicon-based photonics waveguide. Silicon based dielectric material was chosen due to integration capability with common microelectronics silicon-based technology. From literature [3], silicon on doped silicon, silicon germanium, silicon on insulator (SOD and silicon on sapphires (SOS) are first silicon- based photonics waveguides. In that four waveguide systems, silicon on insulator (SOI) become the most popular waveguide system due to SOI technology gives many advantages as lower power dissipation, high radiation tolerance and lower parasitic capacitance. Silicon on insulator (SOD, silicon nitride (SizN,4), porous silicon and silicon crystal are an examples material were used in fabrication of silicon-based waveguide. In that, In process to find surface intensity can be finds by using given equations, De nent: Anwnnitere

Material parameters in this paper as table 1(a).

Fig 2(a). Show layout design in 1-D view.

The refractive index waveguides medium is the mair parameter here and optical input set based on core refractive index. In this studied buried Si3N, core in SiO; cladding with refractive indexs silicon _ nitride (Msi, =1-99) and silicon dioxide (Msjg, =1.46) [12] was used. In fabrication, normally SiOz cladding layer will be fabricate on silicon (Si) substrate layer but due to Si layet has a high refractive index (~3.5), SiOz cladding was used as “substrate later” in simulation methods. In traditional waveguides layers, core layer must be cladding by lower refractive index materials and OptiBPM also only possible to simulate based the rule. Fig. 2 shown Si3N, core buried into SiO2 cladding layout in OptiBPM.

Fig.3. The OptiBPM simulation methods flowchart

Table 1(b), show optical input parameters. Table 1(a), Shown material parameters were used in the simulation.

Fig 4 Surface intensity in various guide layer thickness. The width of waveguide layer fixed at 1.0um. Table 1 (a) show Si3N4 wavegude layer’s parameters and SiO? cladding layer’s parameters were using in the simulation. Due to Si3N, mostly used for thin film fabrication, the thickness set at range 0.luUm to 1.0Um. Table 1(b) is parameter of simulation optical input. Fig.4 is the results for waveguide layer width at 1.0um. By using TE mode polarization, higher surface intensity detected at thickness = 0.1m in fundamental mode. In T mode, higher surface intensity detected at thickness = 0.2uUm in fundamental mode. Both modes (TE, TM) detected more lower surface intensity in second mode ant third mode waves.

Fig. 5 Surface intensities in various guide layer width and guide layer thickness.

Table 1: Taguchi orthogonal Ly array method 2. Literature Review

Vertical Double Diffused Metal Oxide Semiconductor Field Effect Transistor (VWDMOSFET) is the combination of LDMOS and VMOS concept. The design of VDMOSFET power transistor is absorbed with epitaxial layer at various geometrica structures. The characteristics of electrica VDMOSFET power transistor are threshold voltage breakdown voltage and on-resistance. The outcomes o these electrical features are connected to the changes of factors of trench depth, trench width, epitaxia thickness and epitaxial resistivity. The threshold > Figure 1: On-Resistance Transistor Component. a) VDMOSFET and b) UMOSFET (Morancho et al. 1995) voltage 1S Detween ZV tO SV (baliga et al, 1770). Lhe fallen of BV voltage means the voltage is going down and devices flow are transmitted during the gate in close. The barrier between terminal supply and the flow while the gate is in open is known as on- resistance, Ron (William et al, 1998). It is the parameter that controls the value of flowing and spreading power within the devices during it is flow (Locker et al, 1998). According to Chen, VDMOSFET devices, on-resistance as stated Roy = Vps/Ips when the device is closed-circuit. Figure 1 shows all the on- resistance, Roy for VDMOSFET devices including the on-resistance that has channel- resistance, Rg, accumulation-resistance, R,, JFET resistance, Rppr and neck-resistance, Repy.

Simulation for devices of VDMOSFET power transistor is done by writing fabrication recipe and electrical simulation towards the environment. Deckbuild with the ANTENA and ATLAS module are activated and called command go athena and go atlas. Using of Taguchi orthogonal Ly array method, nine models are activated to change the parameter for the trench depth between 0.32 to 0.34um, the trench width between 2.8 to 3 um, epitaxial resistivity is doping 1.0 4. Results and Discussion Result of Device Fabrication

Table 2: The variation of trench width, trench depth, thickness of epitaxial and epitaxial resistivity x 10'° cm* to 1.0 x 10'° cm” and for the thickness of epitaxial between 6 1m to 7 um as shown in Table 2. The test for electrical feature is done on devices n- channel and VDMOSFET power transistor, which is the test of threshold voltage, breakdown voltage and to get on-resistance value. Table 3 shows the Lo array to be insert into variant 4 factor 3 level.

open. Table 5 shows the response value for the breakdown voltage, BV and on-resistance. To get Ron value by using the calculation from equation below, the calculation is done for all devices set. These complete set can be used to gain breakdown voltage and on-resistance value. For the breakdown voltage value, it is the voltage that measure at devices that experience falling down and flowing of electricity during the gate in close. The result of on-resistance is defined as on-resistance that is the barrier between terminal supply with the channel while the gate is

Table 5: Response Value for Breakdown Voltage (BV) and On-Resistance (Ron)

The result from fabrication process to module the n- channel VDMOSFET power transistor device already patterned by Silvaco (1998) is show in Figure 2. Figure 2 shows the cross-section of VDMOSFET was used in first set design of Taguchi method Ly array in Table 3. The first set model of n-channel VDMOSFET power transistor has trench width 2.8 um, the trench depth 0.32 wm, the thickness of epitaxial 6 um and the epitaxial resistivity is doping 1.0 x 10'° cm”. For threshold voltage experiment, the Vyy value is taken Figure 2: Ionization Impact of Breakdown Voltage Table 4: Threshold Voltage Value (V1)

The next analysis for on-resistance respond toward nine experiments of Ly array have been done to determine the required values for selected factors, which are depth, width, thickness of epitaxial, and epitaxial resistivity doping that gave the effect to devices. For the result from calculation of effect factor Taguchi method was shown in Table 8.

After nine experiments of Ly array have been done, the next step is to determine the required values Table 6: S/N Ratio Value for Breakdown Voltage (BV) result. The calculation of S/N ratio for BV was determined by using Eq. | (smallest the better). The overall result of effect factor for breakdown voltage has been shown in Table 7. The S/N value is changing stage four factors with three times changes by using Loy array. This value can be obtained from breakdown voltage value based on fourth factor. The calculation for S/N average ratio breakdown voltage is taken and the average for S/N ratio effect using a specific procedure based on the

Analysis of Effect Factors to On- Resistance Value (Ron)

4.2 Analysis of Effect Factors t Breakdown Voltage Value (BV) for selected factors, which are depth, width, thickness of epitaxial, and epitaxial resistivity doping that gave the effect to devices. Table 6 shows the S/N ratio value for breakdown voltage.

Table 9: Average of effect ratio respond on-resistance (Ron) Optimum Condition for Breakdown Voltage and On-Resistance From Table 8, calculation for average ratio of S/N on- resistance taken from S/N average ratio is according to the specific procedure based on the result table. The calculation of S/N ratio for BV was determined by

Taguchi method suggests the S/N ratio analysis, thus we can get a view concept using graph in order to investigate how the effect for particular factor are examined. From the graph S/N, all factors A, B, C and D are giving effect to the breakdown voltage (BV) with the explanation based on gradient. The optimum Figure 3: Graph S/N for breakdown voltage (BV) and on-resistance (Ron) values are C2 and D1 because S/N value is higher and for factor A and B with factor Al and BO is optimum combination. Figure 3 illustrated S/N graph for breakdown voltage and on-resistance by using Taguchi method. Other method to analize data fr Pareto ANOVA (Park et al ANOVA method were used in ‘om optimum process is . 1996). Principle of this project is same as Pareto principle. Ths work become faster and accurate when using this method to ana ize the result based on obtained data without using ANOVA schedule or F- test. Pareto ANOVA method is suitable for interaction factors condition for analized. Pareto ANOVA for breakdown voltage and on-resistance can be calculated by using value in Table 5. Figure 4 shows the overall result of Pareto ANOVA for on-resistance. breakdown voltage and While for S/N graph for on-resistance, all the factors A, B, C and D are giving effect to on-resistance but has small gradient on factor A and B as shown in Fig. 3. Optimum values are CO and DO because S/N value is higher and for factor A and B with factor AO and B2 is optimum combination. The factor of CO and DO can give more effect because the gradient is high and this means factor CO and DO has bigger related for effect of on-resistance changes.

Fig. 1 An equivalent circuit representing membrane based on Luo-Rudy Phase I model.

Fig.4 A circuit diagram of an analog part in the hardware-implemented cardiac excitation model.

difference which was set up within a resistor, R7, Ix and T,; from Vaigita, and the external impulsive current, [puise from Vp,ise flowed. Finally, they were added up with Iya, Tx;, Ix, and I, from the analog circuit. The membrane potential, V,, changed through the current flow, and the quantity of charge stored in a capacitance, C,,, is equivalent to V,,. Fig. 2 A hardware-implemented cardiac excitation model. Ion currents were implemented by analog and digital circuits.

Fig.5 A circuit diagram of a digital part in the hardware-implemented cardiac excitation model.

Fig. 6 Simulated action potential waveforms. Panels (a) and (b) represent the action potential waveforms generated by the Luo-Rudy phase I model and the hardware-implemented cardiac excitation model, respectively. Panel (c) shows the action potential waveform in the hardware-implemented cardiac excitation model after the scale conversion. Proceeding of National Conference of Electric and Electronic Engineering 2012

Fig.8 A circuit diagram of a impulsive stimulator. Here, as shown in Fig. 8, we constructed an impulsive stimulator by using an H8/3694F microcontroller to control the timing of the stimulation more effectively in realizing the phenomena of reentry. The device was powered by the same source with the digital part. To analyze the agreement between the hardware-implemented cardiac excitation model and the LR1 model, we also carried out numerical simulations of reentrant propagation in the 1D ring of the cable model by using the LR1 model. The simulations were made under the same conditions with those of the active circuit cable of the hardware-implemented cardiac excitation model.

Fig. 7. A ring-shaped cable model. The ring model consists of N cell models and gap junction resistance, R. We are able to obtain a real-time simulation of an anatomical circus reentry around the closed ring active circuit cable consists of eighty analog-digital hybrid active circuits. The first stimulus (S1)-second stimulus (S2) protocol with a impulsive stimulation duration of | ms and an intensity of 150uA was applied to generate the action potential propagation where the S1 is used to pace at a fixed cycle length and the S2 is then applied at an initial interval after the S1. Basically, the stimulations need to be held twice to provoke the reentry, thus the S1-S2 protocol is appropriate as a pacing stimulus for generating the reentrant wave.

itis Known that essential conditions for the initiation and maintenance of the reentry are the development of unidirectional block and the presence of an excitable gap within the front and back of the action potential. It is known that LR model is a modification of the BR model, and the time constants of activation and inactivation of the slow inward calcium current in the BR model are unrealistically long.* Here, both models are modified by dividing time constants of activation and inactivation of the slow inward calcium current I,; with a factor of 12 to obtain the cycle length nearly 40 ms and therefore enable the reentry in 30 cells of the ring cable model. By comparing these two figures, both models show the similar circumstances in the initiation of reentrant wave, although there is a small difference in the range of time between the S1 and S2 stimulus required for generating the reentrant wave. This may be due to small differences in the J-V characteristics between the LR1 model and the hardware-implemented cardiac excitation model, although the J-V characteristics of ionic channels in both models were noted to be generally comparable.

Fig. 10. A space-time diagram showing membrane voltage as a function of time and position around the ring presented by numerical model of the LR1. Fig. 9. A space-time diagram showing membrane voltage as a function of time and position around the ring presented by the hardware-implemented cardiac excitation active cable.

Fig.1 Component 1, Stator of first stage rotary converter. referred in Fig. 1, and Fig. 2 the first stage is designed for operating on the input single phase voltages either 110V or 220 V, at a frequency of 50 or 60 Hz respectively.

Fig.2 Component 2, Dual functions, the inside surface as armature of single phase I.M. and in same time the outside surface as a stator of synchronism R.C.

Fig.4 General Flow chart layout explains sequential steps of design rotary converter electric machine.

3.2 Static Synchronous Compensator (STATCOM)

The capability of STATCOM to have constant current characteristic compared to SVC in situations that voltage doesn’t have suitable stability, it will allows to supplies constant reactive power at the limits, than can be shown in Fig. 4. In voltage regulation mode, STATCOM has the following characteristics:

The phasor diagrams of the STATCOM operation in four distinct quadrants are as shown below in Fig.6. Like any other electrical device, a practical converter

generating or absorbing the reactive power at the point P [1]. The charged capacitor Cg, produces the DC output voltage Wz, to the converter which then produces a set of controllable output three phase voltages V. The voltage V is in synchronism with the AC system voltages performed by an external controller. The controller is actially performing the matching of the voltage !} with ‘;2¢ by controlling the firing angles of the converter. The voltage ¥,,¢ is set manually to the controller. The reactive current injected by the STATCOM into the systeml,, produced by the capacitor with absorbing real power from the AC system. The reactive current is as flown: Ig = — “Teg (5)

Fig.8. UPFC modeling to voltage source structure

Fig. 10. V-I Characteristics of SVC (left) and STATCOM (right) It is quite clear from the V-I characteristics in Fig. 10 that the STATCOM has more capacitive and inductive current range as compared to SVC and is capable of providing more output current than SVC. Moreover STATCOM can provide full capacitive output current at any system voltage, even current with the decreasing system voltage [2]. Additionally, it is evident that the characteristics in transient rating of the STATCOM cover greater regions of capacitive and inductive whereas the transient rating of the SVC is basically reliant on the size of the capacitor [2].also the response of STATCOM in comparison SVC is much faster because STATCOM

Fig. 11. The IEEE 14-bus test system. For analyzing the voltage stability considers a 14-bus test system consists of 5 generators and 11 PQ bus (or load bus) as represents in Fig.11.

Fig.13. Voltage magnitude profile for 14-bus test system without shunt compensation devices. MLP in bus 14 that is known as the weakest bus is 0.69933 P.U.

Fig.15. PV curves for 14-bus test system with Shunt Capacitor As can be seen from the P-V curves, Shunt Capacitor improves the static voltage stability margin of the system. The New maximum loading level in this condition is:

Fig.14. PV curves for 14-bus test system without shunt compensation devices

Table 1. Maximum Loading Point with all types of shunt compensation devices

Fig. 16. PV curves for 14-bus test system with SVC Then, remove the SVC, and insert the STATCOM at the bus14which the lowest the critical point and repeat the simulation. When STATCOM is connected at bus 14, we can observe from Fig. 17 that bus 14 has a flatter voltage profile. The Maximum Loading Point is increasing further at

Fig.18. PV curves for bus 14 with various shunt controllers

It is noticed that bus 5 is the next weakest bus if the STATCOM is introduced at bus 14.

Fig.19. Voltage profile of each bus at the MLP for different shunt controllers

Fig. 2(a) Sequence network connection seen from substation A Fig. 2(b) Sequence network connection seen from substation B

Table 1 Transmission line parameters Fig. 4 Transmission line modelled using Matlab/Simulink Example |:

Using the magnitude of V,, from equation 8 and equation 9, V,, at every from both substations is the intersection point of kilometer until line length seen plotted in Fig. 5. From the plot, voltage from both substations occurs at 15.948 km referred to substation A which means that the estimated fault location is 15.948 km from substation A and the error calculated is 2.107%. Even though the fault resistance for this example is high (40 ohm), the estimated fault ocation is near agreement.

Fig. 7 Errors of fault location estimations for faults along the transmission line Fig. 7 shows the graph of error in percentage versus actual fault location along the transmission line. For any fault location along the transmission line, even though fault resistance is high (Rp = 40 ohm), the error produced is quite similar to fault with resistance, Rp = 5 ohm. It shows that, this technique is not subjected to fault resistance value.

From Fig. 6, it can be seen that the estimated fault location is 34.8277 km from substation A and the error calculated is -0.3666%.

Table 1 Soil resistivity values and the soil model.

Table 2 Tower footing resistance (TFR) measurement.

Fig. 1 Ground stroke density map of the line. The line lightning parameters were obtained from TNB LDS. The GFD is between 10 to 23.3 flashes/km2- year, with average of 16.6 flashes/km2-year. A comprehensive Ground Stroke Density Map for the area surrounding the line is shown in Fig. 1.

Table 4 Simulation results under various CFO and flashover rate

Table 5 Simulation results under various TFR and flashover rate

Table 6 Simulation results under different tower top height (m) and flashover rate, for different applied GFD

Table 6 Simulation results under different % of lightning arrestors installed and flashover rate

Table | Load factors of aggregate loads taken at distribution substations From Table 1, average group load factors for residential, commercial and industrial customers are 0.64, 0.54 and 0.71 respectively. The figures show slight increase as compared to previous year’s measurements. It is therefore recommended to apply updated load factors at the planning stage of distribution system for the respective areas.

Fig. 1 Residential aggregate loads Based on the measurement, load profiles for residential, commercial and small-scaled industrial customers are generically represented in Fig. 1, Fig.2 and Fig. 3 respectively.

Fig. 3 Small-scaled industrial aggregate loads

Fig. 4 Mixture of residential and commercial aggregate loads

Table 2 Coincident factors based on mixed customer load types in Shah Alam Based on load profiles taken from distribution transformers at Shah Alam, Bangi and Subang Jaya, their respective coincident factor is computed. Sample results for Shah Alam area is shown in Table 2. For Bangi and Subang Jaya areas, the coincident factor ranges from 0.65 to 0.9, depending on the load compositions.

Table 3 Average capacity factor of distribution transformers The three areas where the load profiles were obtained are generally matured areas of development. Findings of this study show that a high proportion of distribution transformers are operating with a capacity

Fig. 5 Probability distribution of peak demand of 1000 kVA transformers feeding residential loads Results of the statistical analysis for 1000 kVA transformers feeding residential loads and commercial loads are shown in Fig. 5 and Fig. 6 respectively.

The iron was modeled as a resistor with 52.350 value. Simulation diagram for illegal connection before the meter is depict in Fig.1

The results obtained for power estimated by the RTU, meter and their differences are illustrated in Fig.5a-c respectively. Obviously, the RTU recognized the illegal load while the meter did not, but the illega connection could be tracked as shown in Fig. Sc. Similarly, Fig.6a-b, illustrate the current as observed by the RTU and meter, respectively. The differences (figure 6c) can be used to detect illegal usage of electricity. The voltage graphs for the RTU, meter and differences is show in Fig.7a-b, respectively. The voltage difference is zero (fig.7c) because the conductor is assumed to be perfect. The results of simulated normal condition are shown in Fig.2a-c. The graphs for the power as recognize by the RTU (Fig.2a) and the meter (Fig.2b) are the same, and hence their differences indicate zero current (Fig.2c). The same situation holds for the

currents responses in (Fig.3a-c) and voltages graphs (Fig.4a-c).

Figure 4: Responses of normal connection for voltage on power on RTU (a) Meter (b) and differences (c)

is given as in equation (7) and (8) where P is referring to the predicted value while O, is the observe value. A

The plots of frequency distribution of load profiling data for each day in the week are shown in Fig. 1 to Fig. 7. A study on these plots suggested that the plots for Monday, Tuesday, Wednesday and Thursday are skewed to the left. A unimodal plot is detected for Monday and Tuesday (Fig.1 and Fig.2) while the plot Thursday and Saturd ay display a bimoda for Wednesday, pattern (Fig.3, Fig.4 and Fig.6). In addition, the Friday and Sunday plots show a multimodal on the plots are due consumption for dif! that the plots of the to the different pea be modeled by the se ected three statistica pattern. The different mode patterns k hours of load ferent days. Visually, it is apparent oad consumptions for each day can distributions.

Table 3 Mean square error (MSE) Table 4 Normal absolute error (NAE)

Table 2 Parameter estimates for selected distributions. example of fitted probability distribution function for Monday.

NAE t d d " tables, it can be conclude that the best statistica istri and Saturday is Weibull distribution. Meanwhile the best istri he week. From these value, the selection of best model is Table 3 and Table 4 presents the value of MSE and respectively for each distributions on each day of on the smallest value of MSE and NAE. From the bution for Monday, Tuesday, Wednesday, Thursday bution for Friday and Sunday is log-norma d istri bution.

Fig. 2.1 Final energy demand by sectors on 2010. Source of Data: Ninth Malaysia Plan 2006-2010, Table 19-2. Pusat Tenaga Malaysia (2005). Fig. 2.1 shows that, the percentage of energy demand by each sector in 2010. For sector residential and commercial sectors are the third largest in energy demand compared with others sector as shown as in Fig. 2.1.

For commercial sector, the ownership of electrical appliances are obtained from four classes of commercial sector around Sri Rampai. 25 respondents are chosen. In this sector, it only focused on four primary types of commercial sectors which are hotel, offices/stores, workshop and restaurant. The results of this ownership of electrical appliances as in number of consumers are shown in Fig. 4.2. Fig. 4.2 Ownership of Electrical Appliances in four Typical Commercial Sectors

The ownership of electrical appliances are obtained from three classes of domestic sector around Wangsa Maju and Setapak. 50 respondents are chosen for each residential area. The results of this ownership of electrical appliances as in number of consumers are shown in Fig. 4.1.

The usage pattern of electrical appliances for domestic sector as in consumption of electricity is shown in Fig. 4.3. It shows that the peak hour for using these electrical appliances is between 7.00 am to 9.00 am and from 7.00 pm to 10.00 pm. This data is collected on Friday, 18 November 2011 at one of the sample house at Wangsa Maju and this result shows the usage pattern during workdays. The electrical appliances operated are consumed more than 5 kWh a day.

Besides that, in Fig. 4.4 shows that the peak hour for using these electrical appliances in commercial sector. In commercial sector, this result only focused on workshop sector. The data is collected on Friday, 18 November 2011 at one of the sample of workshop at Sri Rampai. It shows the peak hour for using these electrical appliances is between 10.00 am to 6.00 pm because only at this time, the workshop will operate. At this time, the electrical appliances operated are consumed more than 8 kWh a day. Fig. 4.4 Usage Pattern of Electrical Appliances in Workshop Sector

Table 4.1 Cost for energy usage for three types of residential classes

Table 4.2 Cost of energy use in commercial sector

This test case worked fine in steady state condition until 5 sec, but the system gets disturbed beyond this time until complete and proper LS is applied, because contingency of generator outage is applied at 5 sec. Under these conditions, generator 2 is separated from the system which will cause 163 MW generation oss which is about 50.99% of the total generation. Due to this generation loss f starts decaying as shown in figure 3. In this figure of f plot the hard line shows the use of general (100% static) load while dotted line shows the use f plot when applying 68% dynamic load. The f decays to 47.506 Hz in 0.5218 seconds @ -4.77

Fig. | wind turbine system. Fig. | shows the most important parts wind system Turbine. 3. Power Electronics Devices

Fig. 3 Full bridge inverter topology and its output example. A single phase full bridge inverter circuit and its output example are shown in Fig.3 It consists of four switching elements and it is used in higher power ratings application. The four switches are labelled as $1, S2, S3 and S4. The operations of single phase full bridge converter can be divided into two conditions. Normally the switches S1 and S4 are turned on and kept on for one half period and S2 and S3 are turned off. At this condition, the output voltage across the load is equal to Vdc . When S2 and S3 are turned on, the switches S1 and switches S4 are turned off, then at this time the output voltage is equal to-Vdc. The output voltage will change alternately from positive half period and negative half period.

The power converter is an interface found between the load/generator and the grid. Depending on the topology and the applications present in the system, power can flow into the direction of both the generator and the grid. [4]. In using converters, three important things must be considered: reliability, efficiency, and cost. Fig. 2 shows a single-input single-output power converter system.

In the SPWM method, the fundamental rms output tage in linear region is In unipolar SPWM, two pair reference signals are comparing with carrier signal as shown in Fig. 4. The PWM switch patterns of this technique are,

input neurons. The first input neuron is error signal between desired signal and actual signal. Fig.5 Architecture of the proposed neural network controller The connections weight parameter between Jth and Ly,

Fig. 6 Simulation models of SPWM techniques for single phase full-bridge inverter. The simulation models of SPWM and neural network controller technique for single phase inverter are developed by using Matlab/Simulink. The full-bridge inverter simulates using existing components block such as DC source block and four IGBT block. In order to develop SPWM model, computer programs are designed to generate sinusoidal as references and triangle waveform carrier signals respectively.Using S-Function block, the proposed program of SPWM are uploaded into Matlab-Simulink model as shown in Fig. 6

Fig. 7 Simulation models of Neural Network control for single phase full-bridge inverter. Using S-Function block, the proposed program of Neural Network controller is uploaded into Matlab-Simulink model as shown in Fig. 8.

Fig. 7 Output voltage simulation using SPWM Output voltage simulation result with ma=0.8 of SPWM are shown in Fig. 7.

Fig. 9 Output voltage experimental result using SPWM and Neural Network control . The output voltage experimental result using SPWM and Neural Network control method is shown in Figure 9 .

Fig. 8 Configuration of experimental setup

In the main page there are input selectors for the Lab Jack U3 HV. There are 3 inputs being used in the as AINO, AIN1I, and AIN2. The data section is showing the DC rms value configured by the Lab Jack itself. The STOP button indicates the stop function for the My Project.vi. The Error Massage is being used to display if there is any error between the Lab Jack device and the Lab View software Figure 4.2: My project.vi (Front Panel — Red Phase)

Figure 4.3: My project.vi (Front Panel - Yellow

Figure 4.5: My project.vi (Front Panel—3 Phase current Reading Red, Yellow) Figure 4.6: My project.vi (Front Panel - 3 phases Current Reading Blue)

Figure 4.4: My project.vi (Front Panel - Blue)

On the Red, Yellow and Blue ta b the graph shows t red, yellow and blue phase current value versus t time in seconds. This graph wil amps of the red, yellow and blue phase connected the normal running load the red phase running am are 2.102187 amps. For the yellow phase the running amps are 3.132073 amps. There fore the running am; for blue phase are 1.259409 amps. indicate the running he he to DS ps

achieved is 2.183 from the normal running condition. Therefore the Max Value (Y) graph shows 3.198 amps detected as the highest amps reading for the normal running condition. Meanwhile the Max Value (B) is indicating the highest amps achieved are 1.316 amps from the normal running the data is recorded in the real time system according to the laptop’s time monitoring system. A scroll bar has been interfaced to the graph so that the user can easily back track the back log data on the selected running operation day. Figure 4.7: My project.vi (Front Panel 3 phase accumulated total waveform)

Figure 4.8: Fluke i400s AC current clamp

The 3 phase current rating for red and yellow on the above figure shows the normal running amps with out t d t ( he water heater connected. The Max Value (R) has etect the highest amps on the running conditon is 2.183 amps.. Therefore the Max Value (Y) has detect he highest amps on the running condition is 3.198 amps. Meanwhile on the figure below the Max Value B) has detected 1.396 amps. In this load only fan and flourecent lights are connected. Figure 4.10 : 3 phase current reading with normal running load for Blue Phase

On this analysis a water heater has been connected to the red phase to determine the highest amps captured. The demand graph of the red phase will increase from the running value to the highest value consumed by the red phase, the reminder two phases will be remain same as there are no extra load has been connected. Below are the figures of maximum current detected by Figure 4.9 : 3 phase current reading with normal running load for Red and Yellow Phase

Figure 4.12 : 3 phase current reading with extra running load applied to the red phase (ii) Figure 4.13 : 3 phase accumulated total waveform with extra load connected.

The water heater has been connected to the red phase. The user can compare the difference between figure 4.9 and figure 4.11 where the normal load higest peak deetected is 2.183 amps. Once the load has been applied to the red phase the current have increased to 10.019 amps. Around 7.836 amps have been rise from the normal runing load. The yellow phase has been increased a little due to a florecent light load. It has been up from 3.198 amps to 3.218 amps. While the blue phase as the figure below has not detected any load during this analysis.

Figure 4.11 : 3 phase current reading with extra running load applied to the red phase.(i).

Below are the figure exporting data to Microsoft Excel from the 3 phase accumulated total waveform. Figure 4.14 : 3 phase accumulated total waveform with extra load connected. (Kindly refer to the apendix 4 (table figure 4.14))

The first recorded time of the higest amps is detected and the first amps also will be shown on the Microsoft Excel template. The rest data will be the data without extra load. For more specific dat to be obtain the user can collect the data after the daily operation where the scroll bar acts as the data logger for the whole day. The user can easily scroll to past data record as below figure Figure 4.15 : Maximum demand graph increment for Red phase.

The user can also detect the rising and also the droping of the load graph in rela time and date. From this grpah the user can notify that the selected area in the industrial or in the building is no longer using the load. The user now can decide wherever to start the equipmet again due to the highest amps already recorded. Below are the figure of the data exported to Microsoft Excel. Figure 4.18 : Maximum demand report decrement in Microsoft Excel for Red phase.

Figure 4.16 : Maximum demand report increment report in Microsoft Excel for Red phase. (Kindly refer to the apendix 4 (table figure 4.14)) Figure 4.17 : Maximum demand graph decrement in Microsoft Excel for Red phase.

Analysis and Monitoring of Maximum Demand Penalties for Low Voltage Room Power Consumption Methodology flow chart of the Project Appendix 2

Figure 4.1: My project.vi (Block Diagram — 1* part)

This part will control the data from an index array and sent the data to the multiply and to the dc rms block. In this pa also there the max value has been detected in the Array max & min .vi.

Figure 4.11: My project.vi (Block Diagram — 3™ part) In this part the data collected after processing will be visualize in the graph and waveform indicator in the front panel.

This part contributes the data collection of the data for the demand graph.

This part the data collection for the total waveform graph.

This part contributes the Stop button function and the stop function for the while loop case.

Figure 1: Phases of a three-phase system [6]

Figure 3: Phases of a six-phase system [8] Comparing the insulation clearances with the voltage (neglecting the tower width) it observe that bear no relation to each other. Thus, spatial distribution of insulation clearances do not match the voltages experience by them. Therefore, the line is not optimized to this extent.

Figure 2: Double-circuit three-phase line [7] Generally, a typical balanced three-phase system has 120° electrical degrees apart between each phase, such as shown in Figure 1. Based on this Figure 1, the relation of phase-to-phase voltage and phase-to-neutral voltage could be obtained. The phase-to-phase voltage is V3 of the phase-to-neutral voltage. In a three phase system, the three voltages with respect to ground can be written as the phasors.

In contrast with 120° phase angle of the three-phase system, a balanced six-phase system has 60° electrical degrees between each phase as shown in Figure 3. From this figure, we can obtain the relation of phase-to phase voltage and phase-to-neutral voltage for a six-phase system. Figure 4: Single-circuit six-phase line [7]

Table 1: Lightning and switching overvoltages according to IEC 60071 [12]

Overvoltage types are usually classified according to their waveform shape, duration and magnitudes. Figure 5 below shows typical overvoltage magnitudes and duration [9, 10], it classifies them according to the IEC 60071-4 standard [10].

Figure 6: Creepage length, insulator length and insulator length with fittings as observed on insulator The length of the insulator must provide sufficient clearance to prevent flashover in the air caused by transient overvoltages. A typically 275 kV insulator designed for an L3 tower has a total length of about 2.75 m, such as illustrated in Figure 6.

Table 6: Statutory minimum clearances to ground level [22] All aluminium alloy conductor (AAAC) type conductors have also been specified by [23] as the preferred choice for all new conductors erected on any NGC overhead line towers. Table 7 shows the standard type of AAAC conductor bundles that are typically being used on NGC’s 275 kV L3 overhead line towers [23].

Table 7: Preferred AAAC type conductor bundle for NGC’s overhead line towers [23] Using the method and information presented, it is possible to initially determine the physical dimensions of an overhead line. Some formulae have been developed to calculate the clearances between the conductor bundle and nearby objects (e.g. cross-arm

Table 8: Cross-arm widths as observed for NGC overhead line towers

Figure 7: Safety distance as applied to an overhead line [11]

The statutory ground clearance, the standard insulator length, the length of conductor fittings and the maximum sag of the conductor bundle can be used to calculate the height of the bottom cross-arm of the tower. For example, a bottom cross-arm standard 275 kV L3 overhead line tower with 11.06 m maximum conductor sag is calculated to have a minimum height of 21.9 m on a normal open field at ground level (7.6 m statutory ground clearance + 3.25 m insulators with fittings +11.06 m sag). The method is illustrated in Figure 8. Figure 8: Bottom cross-arm height for 275 kV L3 tower

Figure 9: Cross-arm widths for 275 kV L3 tower The vertical spacing between two cross-arm on a overhead line tower is defined by the length of the insulator and the required clearances, both when the conductors and insulators are still and when being swung by wind load.

Table 9: Cross-arm spacings as observed for NGC’s overhead line towers Figure 10: Cross-arm spacing for 275 kV L3 tower Table 9 shows the example of the spacing for top, middle and bottom cross-arms for the standard NGC’s 275 kV L3 overhead line towers. Figure 10 illustrates the method with some spacing examples as observed on the former type of tower.

As for work described, the ST_cateuiatea Values were obtained by using the necessary positive polarity switching USO formulae and the gap factor values. Figure 11: Necessary software (Microsoft Excel) were used to conduct the analysis

Figure 12: Clearance as a function of insulator length for the analysis of three-phase system tower design

Fig. 1 Activated Sludge Reactor The nonlinear biological wastewater treatment process which has been studied in this paper is taken from [3]. The substances from the substrate (S), represents a source of energy and carbon (necessary for the synthesis of new cellular material) that are used by the microorganism (biomass — X) to live and reproduce

where the number of column j is assumed to go to infinity throughout this thesis. The subscripts of U and Y denote the subscript of the first and last element of the first column. The past inputs and future inputs are denote as Upyz-a =Uz and Uyo;_,= Uy respectively. A similar notation applies for the past and future outputs. This notational convention is useful when explaining concepts. Input and output block Hankel matrices are defined

Fig. 2 Subspace and classical methods of system identification

Fig. 4 Principle of receding horizon control strategy Model Predictive Control or MPC, is an advanced method of process control that has been in use in the process industries since the 1980s. Model predictive controllers rely on dynamic models of the process, most often linear empirical models obtained by system identification. Fig 4 shows the receding horizon control strategy used in MPC. Although this approach is not optimal, in practice it has given very good results. Much academic researches has been done to find fast methods of solution of Euler- Lagrange type equations, to understand the global stability properties of MPC's local optimization, and in general to improve the MPC method. To some extent the theoreticians have been trying to catch up with the control engineers when it comes to MPC.

MPC control design. The design of adaptive MPC can be performed in two different ways; indirect method and direct method. Fig 3 shows the differences between the two methods. Fig. 3 Comparison between indirect (left) and direct (right) schemes for obtaining controller parameters This study will only focus on the direct approach. Note that, in the direct method the steps of identification and control design can be carried out simultaneously by applying a single QR decomposition to the input-output data. This stands in contrast to the design of indirect adaptive MPC.

The computation is more involved for the constrained case, particularly in the multivariable constrained systems. The problem takes the form of a standard quadratic programming which can be solved online. Since the DAMPC uses sliding window method, the solution to the QP problem leads to a highly computational burden as the size of identification window increases. Fig 5 represents the block diagram of a subspace based adaptive MPC. In DAMPC, the controller parameters can be retrieved directly from the identification steps and the possibility of handling constraint implicitly is explored. To circumvent this problem, we consider the data-driven DAMPC control strategy for increment input constrained multivariable system. The strategy is based on analyzing

Fig. 6 Simulink model used to implement SPC on a wastewater treatment plant system A simulink model of the implemented subspace predictive controller was created to control a stable MIMO system that represents the wastewater treatment plant. The model used is shown in Fig 6, and the SPC controller subsystem can be seen in Fig 7.

The system used was the same linear model obtained previously. In the simulation, the following parameters were used:

From Fig 8, it was found that when the dynamics were changed during the simulation from the original system to a new value of weighting matrix (Q = 4 and R = 1) the system responded well, with the control continuing to stabilize the system before the identification had collected a new set of input-output data and calculated new control matrices. Fig 9 shows the same results but for t=400.

Fig. 1 .Above figure shows the proposed Master (Smart Glove)and Slave (artificial hand gripper).

Fig. 2;Above figure shows the proposed Bluetooth system and Data logging in computer system.

3.1 Flex sensor Fig3. Bluetooth Channel Analysis for master and slave Due to this projects requirement, it needs two unit of Bluetooth module which is different personal add ress each other’s. The address identification setting for master and slave module is same but in differerent address. User can decide which module unit is Master or Slave. Add for master Bluetooth unit that state by user 646e6cO0beeb and for slave is 646e6cO00ab56. In ress 1s this experiment, user state that slave unit of Bluetooth is 646e6c00ab42, but the hardware problem happen due to the Bluetooth module, user replace with ano ther Bluetooth module that same type but different Address. As show on figure 3.

Fig. 4. A Flex sensors attached on the back of Smart Glove.

Graph 2 and 3 shows the results of experiment done on the flex sensor. We applied bending on the flexi sensor active surface 5 times, and the graph shows 5 peaks which suggest that the sensor can be used to detect bend when the glove bending 90°, 60°,45° and 30° . By attaching this sensor to the gripper fingertips, the flex sensor will act as a detector which sends data to the microcontroller to inform about the gripper is grasping an object. We expected to be able to control the amount of deferent object generated on the grasped object based on the resistant value generated from the flex sensors. Graph 2: Flex sensor bending angle during experiment Degree vs Voltage

Fig5 : Picture above showing the activity of Artificial hand Gripper (1) bend 0° (2) bend 30° (3) bend 45°(4) bend 90°

Graph 3: Flex sensor bending angle during experiment Degree vs Resistance

Fig. 6. Preliminary experiment on determining characteristic Bluetooth

The robotics localization and mapping problem has gained researcher's attention regarding its capability to support for autonomous robot for more than two decades. It states a problem that illustrates a mobile robots observing environment and collecting information efficiently while they are moving through the environment. From the observation, robot makes estimation about the map from what it believes to be. This problem is popularly known as ‘chicken and egg' problem, and there still remains a lot of tasks to be solved even though a rapid progress have been achieved lately. Since 1990's, after a series of influential seminal papers by Smith and Cheeseman et.al[1], these research findings have directly impact to the robotic mapping research and finally has evolved its name to Simultaneous Localization and Mapping problem(SLAM). SLAM is also known as Concurrent Mapping and Localization (CML)[2]. See Figure 1 for further explanation. Fig.1 Simple SLAM illustration Nowadays, SLAM has been applied in various applications, indoor or outdoor such as satellite, mining, space exploration, rescue, and military. The development of SLAM continues whether in 2D[3Jor 3D applications[4][5][6] and has amazingly expands even to

To evaluate H,, Filter performance, the robot is defined to be more confidence about its location with small uncertainties. We manage the simulation longer about 10000s to determine the stability and consistency between these filters and show the filter eligibility towards SLAM. Fig. 2 The global coordinate system representing the location for robot and 2 landmarks to be estimate.

Fig.3 Robot position estimation between Kalman Filte: and H,, Filter On other hand, if y is not selected properly, the estimation of H,, Filter is diverging and subsequently affecting the competency of H., Filter towards estimation. See Fig.9 for details illustration for the effect of bigger observation noise e.g observation noise, R=10. At the beginning, the estimation is rather the same as Kalman Filter. However, the attenuation becomes bigger and goes far apart from the true or expected values as time passed by. Of course, for bigger values of observation noise than that observed before, it is expected that it will excessively affect the inference result and causing H., Filter may not suitable to use for estimation. As stated previously, a bigger value of y which bigger than observation noise, the characteristics will be approximating Kalman Filter characteristics.

Fig. 4 Detail Differences between filter performance for robot estimation

om ee ee Fig. 7 Detail difference between filter performances for landmarks estimation

Fig.5 MSE performance of robot estimation

Fig.6 Filter performance for landmarks estimation

Fig. 9 Effect of R>y of landmarks estimation

Fig.8 Uncertainties are decreasing for all observed landmarks.

Fig.11 Robot location estimation about x coordinate

Fig. 10 Stationary robot observing landmarks

Based on these result, it is indeed shows consistency with the results obtained previously[3] with a slight improvement from H., Filter. Belong to this results of fast convergence, process time for SLAM may reduce significantly and definitely nurture the SLAM problem. These results inspire further achievement and development of H., Filter.

Figure 1: Flowchart of the system Based on the flowchart of figure 1, development of this research starting with data gathering from experimental test by using FES for electrical stimulation test, hence pendulum test for knee joint experimental data by using Goniometer [17-19]. Data collection from anthropometric data is the data parameter of lower limb characteristic.

Figure 3: Application of Electrical Stimulator with surface electrodes

Figure 6: Equipment for experimental setup Due to SCI problem that illustrated in figure 5, pulse signals from brain cannot go through the human body system because of the signal have blocked. Then every level of injury base on SCI problem divided into C6,C4,T3 and L1 .In this research area, there are focused on lower limb paralysed injury of paraplegia and data will be collected during this research development study [7]. Based on the experiment from the FES architecture in figure 2, there are need some another supporting equipments during this research which is to achieve the goal in this project. Some kind of measuring meter such as Digital Multimeter and oscilloscope will be used with signal generator to generate pulse in order to compare pulses with FES (Hasomed GmbH) shown in figure 6.

Figure 5: Effects of SCI depend on the type and level of the injury

Figure 7: Recumbent tricycle for spinal cord injured persons with Functional Electrical Stimulation (FES) of the paralysed muscles with motor assist [22]

Fig. 1a: Illustration for BCI-FES system

Fig. 1b: Proposed Flowchart of the system In order to translate the human’s intention to other devices, BCI is needed which is a kind of communication tool that communicate between human and devices. The entire signals that produced by brain will be recorded and filtered by using this BCI system. The system will

Fig. 3: Architecture of BCI-FES System ~~ —_"- ee Referred to SCI problem that illustrated in Fig 2, pulse signals from brain cannot go through the human body system because of the signal from brain to muscle has been blocked. Therefore, to make the brain signal transfer to the muscle, integration with the FES is needed. In this study, the FES will be used. Also, there are only focused on lower limb paralysed injury of one paraplegic person.

diagnose, record and generate data from human brain activity by using Matlab Simulink.

Proceeding of National Conference on Electrical and Electronic Engineering 2012 Signal Representation Wavelet transform is known as a flexible approach for signal decomposition because various wavelet functions can be chosen as Mother Wavelet Function to analyze the signal [13]. Furthermore, the user can design a wavelet function if it is needed. Khezri and Jahed created a new multi-wavelet function for SEMG processing to recognize hand movements. They also used two intelligent classifiers, artificial neural network and fuzzy inference system to decrease the effect of classification technique on the accuracy of the system. Finally, a multitude of mother wavelet was selected by evaluating some wavelet such as Haar, Daubechies, coiflet, Symmlet and biorthogonal[14]. There is no rule for choosing a wavelet function in a particular application. Sometimes, more similarity between the shape of the wavelet function and the signal can be a suitable criterion for selecting the mother wavelet. But in some other is reported in [16] by studying a huge number of potential mother wavelet functions. The comparing criterion for choosing mother wavelet function was the outcome of the correlation between the signal and wavelet basis function and mother wavelet from different families had been analyzed. Eventually, the results showed that db44 from Daubechies family is the most similar function to study three biomedical signals, EMG, EEG, VPA [16].

Regarding the limitations of EMG control system, such as time delay because of huge amount of data to be processed and consequently considerable computational cost of the control system, Fuzzy Wavelet Packet (FWP) based Feature Extraction was offered by R. N. Khushaba and A. Al-Jumaily [26]. Firstly, a wavelet tree was produced by WPT through decomposition stage using a Symmlet family with four levels. Afterwards, non- distinguishing features are removed by using Fuzzy C- Mean (FCM). Next, PCA was employed for feature dimensionality reduction and lastly, a multilayer perceptron neural network classifier accomplished the classification process. Therefore, the results showed using only the first 20 principle components obtained an accuracy of 99% for six subjects. After this research, R. N. Khushaba, A. Al-Jumaily and A. Al-ani discovered another novel feature extraction method based on Fuzzy Entropy and WPT [27]. In this way, WPT with four leve of decomposition was applied on EMG signals and A combination of Wavelet theory and AI methods was reported by Y. Kocyigit and M. Korurek using WT and Fuzzy clustering Algorithm in 2001 [21]. They extracted Feature vectors by applying DWT but because of high dimension of the feature vectors, the dimensionality was reduced by employing Principle Component Analysis (PCA) method since according to K. Englehart, B. Hudgins and P.A. Parker [22], [23] applying PCA is ssential to improve WT feature extraction stage. Fuzzy ustering algorithm in [21] included Fuzzy C-Mean (FCM), Possibilitic C-Mean (PCM) and Fuzzy K-Nearest Neighbor classifier (FKNN). Eventually, a Daublet mother wavelet of order six resulted better accuracy than other wavelets with PCA, and PCM demonstrated more classification improvement in this project than FCM. Wavelet based neuro-fuzzy classification was proposed for EMG control in 2002 [8]. In this regard, DWT with Mallat algorithm was utilized and for initialization of fuzzy membership function of Neuro-Fuzzy Classifier (NFC), C-mean clustering algorithm was used and rules

Figure 2: GSM Controller board schematic diagram Figure 1: Identified of innovated system 4. Experimental and Results

Figure 3: Communication between mobile phone and GSM Controller Communication occurs when monitoring message on the car sent by mobile phone to GSM modem which was installed inthe car compartment. SIM card inserted in the GSM modem acts as a receiver at the same time also serves to send back the car status report in the form ofa messageto the users’ mobile phone as illustrated in Figure 3 and Figure 4 is shown the process flow chart of the developed system.

Figure 4: Process flow chart of the system These are the procedures to test and trial run process of 4 innovated systems based on 5 situations as following:

5. Discussions The system that has been developed with unique features such as the following:

Fig. 1 System Features of proposed system The system requirement is based on inputs and output of the system features. Fig. 1 shows the system features that are divided into two main categories, which are data collection unit and also the display function.

Fig. 2 System Architecture seven LEDs. All of this hardware is controlled by a microcontroller PIC18F4550.

Fig. 5 Flow chart of medium-level software " [he medium-level software is a subroutine whic combines the sensor output and also the timing delay fo the LED to be switched on and off. Timing delay is set a the beginning. As the IR sensor input HIGH, it will kee detecting. As the IR sensor input LOW, it will outpu HIGH to the microcontroller. Then microcontroller wi send signals to LED to be switched on and o accordingly in desired manners. Next, the lighting manner is supported by timing delay of 0.76ms for every 1x7 pixel as shown in Fig. 5. rR PY ao a

Fig. 6 Flow chart of high-level software To achieve and satisfy with the requirement of the proposed system, several testing have been conducted. Firstly, the fan’s AC motor speed was tested by using tachometer equipment and the data in rpm was recorded. Then, timing delay is calculated based on the speed provided by the motor. Before assembling on the prototype of the proposed system, the relationship between speed and timing delay were determined via Proteus’s simulation to obtain the output waveform of the voltage output in time in oscilloscope. Lastly, the program was assembled into the prototype and it was tested to obtain the output waveform in real time oscilloscope. 3. Results and Discussion

Before using the table fan AC motor as the hardware for this proposed system, few tests were conducted on 24V DC brushless Vexta motor and the 12V starter motor from Modenas brand. speed at 300 rpm and However, Vexta motor offers low Modenas starter motor consumed power very fast and lack of stability although high speed of rotation is provided . Table fan AC motor offers the most suitable speed and moderate consumption of power. Therefore, it has been chosen to be the main rotary hardware in this proposed system.

then being regulated down to 5V for powering controller board. The general image of the proposed system has been shown in Fig. 7 and the hardware description data has been shown in Table 2.

Fig. 8 Image of spinning LED message After getting the accurate result of readings for the proposed system, IR sensor is added into the system as position sensor. It is used to avoid the drifting of the message and fix the message starting at the same place. The sensing distance is adjusted manually. The image of the proposed system displaying LED message is shown in Fig. 8. Other alphanumerical characters were also tested during the project duration such as “2011”.

The system architecture of differential wheeled mobile robot is shown in Fig. 1. This robot is controlled by a microcontroller PIC18F452. It has two DC motors for driving the movement that are attached to left and right wheels. An ultrasonic range finder sensor is connected to analog pin of microcontroller. This sensor is used to measure the wall distance. There are five IR digital range sensor have been used to keep following the wall and avoid obstacle. The robot has square shape, therefore, four IR sensors are located at edge of the body corner and another one IR sensor located at front. To display the robot status and sensory data, a LCD has been used and its display change based on user input through two push buttons. Besides, to create more tasks, the microcontroller needs to control seven LEDs alongside other tasks such as motor control and sensory data reading. The detail main components used are specified in Table 1.

The higher number defined higher priority of the task. The tasks that need to be update more frequent such as the first task (motor control) had to assign with the lowest priority so that it only activate the task when other tasks is in waiting state. If the first task is assigned with highest priority, the program will only execute the first task, because the kernel is busy to update the first task for every 5ms and as a result, the other with lowest priority is suspended. In Fig. 3, it shows the running LED has been running and the LCD displayed message” Press SW1 start”. Next snapshoot shows the IR sensor detected an obstacle at front, the ultrasonic range sensor will start measure the distance and the LCD will update the distance value for every 200ms meanwhile motor change its direction to backward. At the same time the running LED is updated every100ms. Next, the system shows the IR sensor detected an obstacle at left, after that, the LCD display the message “Twist CW” and the motor change its direction turn right and the running LED is updated. Same thing happened in next snapshoot, when IR sensor detected an obstacle at right, the LCD display the message “Twist CCW” and the motor change its direction turn left and the running LED is updated.

Table 3 Testing With and Without Using PICos18 4.3 High Level

Fig. 1 Dead reckoning algorithm travelling path In Fig. 1, a dead reckoning algorithm travelling path is shown, where T, are the encoder ticks recorded on left drive; T, are the encoder ticks recorded on right drive; Ry is the radius of each drive wheel; Tp is the number of encoder ticks recorded in a full, in place rotation.

Fig. 2 PID line follower algorithm travelling path One idea in controlling a line following robot is using the PID control algorithm [4]. Most importantly, PID control uses continuous function to compute the motor speeds, so that the jerkiness of the previous condition can be replaced by a smooth response. The terms of Proportional-Integral-Derivative are described as three input values used in a simple formula to compute the speed of robot either in turning left or right. The PID line follower algorithm travelling path is shown in Fig. 2.

Fig. 5 Components on the bottom layer PCB The bottom layer as shown in Fig. 5 is an IO interfacing circuit board. This is because all inputs sensors or output controller pins are connected on this board. Since this bottom layer PCB is designed without any microcontroller on it, a top layer PCB is required in case.

Fig. 7 Dead reckoning for straight line travelling concept The concepts to design dead reckoning of straight line travelling and dead reckoning of curving travelling concept are shown in Fig. 7 and Fig. 8 respectively.

Fig. 6 Block Diagram for Dead Reckoning Algorithm The odometry data or encoder reading will be collected as reference. The block diagram for dead reckoning is shown in Fig. 6.

Fig. 8 Dead reckoning for curving travelling concept

Table 3 Predictable Digital Sensors Position and Errors

Fig. 2 Block diagram in term of PID algorithm Fig. 1 Block diagram for PID line follower algorithm

Fig. 9 The arrangement of sensors’ position for PID line follower

Fig. 3 Motion planning system block diagram It is declared as the final designing stage in this project. The overview of motion planning system flowchart is shown in Fig. 12.

Fig. 4 Straight travelling path of autonomous robot In the straight line travelling path observation, the autonomous mobile robot moves a straight line path, from the position A to the position A’, with the heading feedback of digital compass. Then, the mobile robot stops for a desired distance which is measured in 1400mm. The travelling path is shown in Fig. 13.

Table 6 Heading Results of Curve Travelling Path

Table 7 Odometry Results for Curve Travelling

Table 5 Odometry Results for Straight Travelling

Table 4 Heading Results of Straight Travelling Path

Fig.5 Curve Travelling Path of Autonomous Robot

Fig. 6 Robot Responses for Proportional (P) Controller

Fig. 16 Robot Responses for Proportional-Integral (PI) Controller

Nevertheless, there are three values for D parameter used, including D = 12, D = 18, and D = 24. In this experiment, the robot responses for PD controller are observed and plotted in the graph, as shown in Fig. 17.

Fig. 188 Robot Responses for Proportional-Integral- Derivative (PID) Controller would be measured by using the digital compass module. The motion planning system was implied on the robot in finding the shortest distance to its goal. First, robot had been placed in the game field of coordination (0, 0). It would be prompt in a desired coordination by programming; the desired coordination was set as (-1, 3) in case. The position of robot and its coordination in the game field were demonstrated in Fig. 19. According to the motion planning system concept, the autonomous robot was able to travel in the shortest travelling path in order to reach to the desired coordination. The robot movement result was demonstrated in Fig. 20.

Fig. 9 Robot's Initial Coordination & Desired Coordination of Case A

Fig. 20 Robot's Final Coordination & Its Travelling Path of Case A

Fig. 21 Robot's Initial Coordination & Desired Coordination of Case B 21. Since the initial heading direction of Case B was differ in comparing with that of the Case A, the robot could navigate to find a new travelling path in order to reach it desired goal. The travelling path of the robot Case B would be demonstrated in Fig. 22.

Fig. 11 Robot's Initial Coordination & Desired Coordination of Case C In the Case C analysis, the motion planning system would be proved that the autonomous mobile robot able to travel by combining the two types of algorithms, including PID line follower algorithm and dead reckoning algorithm. Same as last previous two cases, the autonomous robot had been placed at its initial coordination (0, 0). For this situation, there was a new desired coordination, which was (-2, 4), had been prompt in the motion planning system by programming. The position of robot and its coordination in the game field was demonstrated in Fig. 23. Based on the given situation, the robot’s travelling path of Case C was demonstrated in Fig. 24.

Fig. 10 Robot's Final Coordination & Its Travelling Path of Case B References

Fig. 24 Robot's Final Coordination & Its Travelling Path of Case C

An accelerometer is a device to measure the linear acceleration of the device itself. The measurement unit is often reported in the unit of the gravity or g-force. The working principle is based on Newton’s Laws and could be illustrated using a system of a proof mass and a spring as shown in Fig.1. The force (f) is an inertial force when the accelerometer is accelerated. According to Hooke’s Law, the spring displacement (d) while the spring is working within the elastic limit is proportional to the force applied onto it. The total force (F = ma) acting onto the mass as shown in equation (1)is the sum of the inertial force (mf) and the weight of the mass (mg) under the effect of the gravity. During static or moving in a constant linear speed, the inertial force (f) will be zero leaving only the gravitational term in the equation.

The accelerometer will be calibrated through these procedures. The rotary table is turned manually in counter clockwise (CCW) direction. From 0°to 360°, stepping at 1° interval, the IMU outputs (X, Y, Z, XR, YR) are recorded. The data are saved in a secure digital (SD) card for post-processing. Multiple experiments are repeated in every plane. Fig.3: The test rig of the rotary table

A positive g-component means that the accelerometer axis is pointing against the gravity and vice versa. For example, at position-1, X and Z axes are tilted at 45° upward, their outputs are +1g-sin(45°) = 0.7071g while Y-axis is at Og. These measured outputs displayed very high repeatability as the maximum standard deviation of measured outputs at 4.7 x 107-°V_ was observed, suggesting negligible noises for all collected data. Fig.4: The test rig of 6/12 Known Positions

Fig.6: Setup of the in-use calibration The experimental setup for an in-use calibration is shown in Fig.6. The test rig consists of an IMU mounted in a cube and a customized datalogger. The cube is swung randomly and rotated freely in air for a few seconds. The motion is random but slow. All IMU-axes outputs are recorded. IMU1 to IMU6 are experimented in the same method.

Table 1:Comparison of static calibration techniques

Key takeaways

Various application are using according to specific demand like fast devices that use high voltage.
The comparison show that by using Neural Network controls the output voltage can be more stability .
This analysis and monitoring of maximum demand for low voltage room power consumption has been build based on the literature have been made.
Voltage is a potential energy source in an electrical circuit that makes things happens.
The fact that power is a part of voltage and a part of current and if the power is zero then both voltage and current is zero.

Log In

Direct Adaptive Predictive Control for Wastewater Treatment Plant

Sign up for access to the world's latest research

Abstract

Figures (490)

Key takeaways

Related papers

Related papers