International Review of Mechanical Engineering (IREME

Paul Kah

International Review of Mechanical Engineering (IREME

Abstract

This paper presents the results of experimental studies to control the base pressure from a convergent nozzle to ascertain the effect of level of expansion on a suddenly expanded sonic flow. An active control in the form of four micro jets of 1 mm orifice diameter located at 90 0 intervals along a pitch circle diameter of 1.3 times the nozzle exit diameter in the base region was employed to control the base pressure. The area ratio (ratio of area of suddenly expanded duct to nozzle exit area) studied is 2.4. Experiments were conducted for nozzle pressure ratio (NPR) from 1.5 to 3, in steps of 0.5. The length-to-diameter (L/D) ratio of the enlarged duct was varied from 10 to 1, and tests were conducted for L /D 10, 8, 6, 5, 4, 3, 2 and 1. It is evident from the results that the flow field downstream of the reattachment-redevelopment is very complex. It is found that, unlike in the case passive controls, the favourable pressure gradient does not ensure augmentation of the control effectiveness for active control in the form of micro jets. To study the influence of micro jets on the quality of flow in the enlarged duct wall pressure was measured and it is found that the micro jets do not disturb the flow field in the duct.

Figures (457)

Fig. 1. Sudden Expansion flow field The point at which the dividing streamline strikes the wall is called the reattachment point. In such a case the shear layer divides such a flow in to two main regions, one being the re-circulation region and the other the main flow region (Fig. 1). In view of above it has been the subject of intensive study for many years due to its academic interest and its real-world applications.

It has 16 channels and pressure range is 0-300 psi. It displays the reading after averaging 250 samples per second. To interface the transducer with the computer, software is provided by the manufacturer. The user friendly menu driven software obtains data and displays the pressure readings from all the 16 channels simultaneously on the computer screen.

Fig. 3. Percentage change in base pressure variation with L/D

International Review of Mechanical Engineering, Vol. 8, N. 1

The wall pressure distribution for this area ratio 2.4 is shown in Figs. 8 to 22. It is seen that when control is activated the quality of the flow in the enlarged duct is not disturbed.

Fig. 1. Schematic diagram of the flexible manipulator system Fig. 1 shows a schematic representation of the flexible manipulator that can be modelled as pinned-free flexible beam. The pinned end of flexible beam of length L with inertia I, is attached to the rotational axis of the hub driven by an input torque r(t) and payload mass M, is attached at the free end. FE, I and p represent the Young Modulus, second moment of inertia and mass density per unit length of the flexible manipulator respectively.

The fourth-order partial differential equation (PDE) in Eq. (2) represents the dynamic equation describing the motion of the flexible manipulator. Note that the model in Eq. (2) does not incorporate damping. The well-known goveming equation of the manipulator with the associated boundary and initial conditions can thus be obtained as [13], [14]:

A is a constant n x n matrix whose depend on the flexible manipulator specification and the number of sections the manipulator is divided into, as Therefore, the required relation § algorithm characterizing the behavior of the flexible manipulator structure is obtained as in Eq. (6), which can be simulated easily.

Fig. 2. Finite Difference discretisation of the manipulator in distance and time of the PDE in Eq. (2). The FD method computes solutions of the model only at discrete points. This involves discretised the manipulator into n-equal length sections, each of length Ax, and the manipulator motion (displacement) of each section at sample times At as shown in Fig. 2. In this respect, let w(x,t) be denoted by Ww, representing the manipulator displacement at gnd point i at time step j. and 2, r) w(x,t) /at? can be approximated as Eq. (4):

A bang-bang torque input with an amplitude of +0.3 Nm and duration of 0.5 seconds is applied at the hub of the manipulator throughout this simulation as shown in Fig. 3. This type of input is chosen to ensure that manipulator link is move accordingly to the input torque.

Figs. 4 show the simulated output of end-point displacement in time and frequency domain. The average end-point displacement is about 0.33m. The first vibration modes are found to be 11.65 Hz with 7.54% of error as shown in Table II. It is noted that the behavior of the manipulator is characterized by oscillatory response where the maximum displacement occurred at the end- point of the manipulator. Copyright © 2014 Praise Worthy Prize S.r.L. - All rights reserved

Fig. 6. Flowchart of Genetic Algorithm Genetic algorithms consist of a population of individuals that each represents a possible solution to the optimization method. This population is then used to generate a new population based on genetic operators such selection, crossover and mutation that will compete for survival leading to the optimum solution. GA perform parallel search for an optimal solution and thus able to find the global minima However, GA typically take longer time to find the solution due to the computationally effort.

In this work, autoregressive (ARX) structure is chosen to model the system that can be described as:

The optimization of PSO is in a guided and intelligent manner and can perform complex tasks as a group [18]. Stages of PSO can be comprised as follows: Xq(t) and vq(t) represent position and velocity

Figs. 9. Swarm movement where (a) Step 1; (b)-(£) Step 2-4

Figs. 11. End point displacement modelling using PSO (b) Err between actual and predicted output

This research is supported using UTM Research University grant, Vote No. 04H17.

The conelation tests including autocorrelation of the error and cross correlation of input-error, as illustrated in Figs. 12 and 13 for both GA and PSO optimization method. The corresponding correlation tests for identification using GA and PSO were found to be within 95% confidence interval indicating adequate model fit.

In loss function indicates the better performance by performance characteristics. Better performance in this experiment is lower response (surface roughness). Hence, smaller values of feed rate and depth of cut must be selected in order to achieve better surface finish during tuming of stainless steel. The loss function is calculated the deviation between experimental values and the desired values. Then, it’s transferred into a signal to noise S/N ratio. The S/N ratio is calculated as follows: Si i a After ran 12 experiments, the surface roughness readings are used to optimize the surface roughness. Generally, increasing of cutting speed caused the larger surface roughness.

Data analyze by Analysis. of Variance (ANOVA) where the relative percentage contribution of all factors is determined by comparing with the relative variance. Table V shows results of Response and S/N ratio for each cutting fluids. In addition, Ra is filled array in this includes the Signal-to Noise (S/N) ratio of the individual mins. The y; is the individual surface roughness measurements in a run for both noise conditions while n is the number of replications. In this case, nis equal to 1.

THE RESPONSE OF BEST COMBINATION OF PARAMETER Based on the confirmation run, we find the minimum esponse for surface roughness at each cutting fluid. Te: whew Phx ‘honed wre’ pawns et” eeeibbess 21.7) Se Table VIII shows the response of confirmation run. It shows the best performance of cutting fluid is Soybean Oil.

ae eS aera THE RESPONSE OF BEST COMBINATION OF PARAMETER Table XVII shows the best optimal cutting fluid for tool wear is Sunflower Oil. It shows the lowest reading of tool wear from the confirmation run of the experiment as compared to the results in Tables XVITI-XIX.

Fig. 3 compares the ability of the mentioned materials to absorb the impact energy in terms of Specific Energy Absorption, SEA. In fact, the parameter of SEA equals to the absorbed intemal energy over the mass of the structure.

Fig. 1. Side door beam location in the frontal vehicle door

Fig. 3. Comparison of SEA among three different SDB matenals In addition, this factor is defined exclusively for crash problems in order to achieve the maximum possible absorbed energy together with the minimum weight.

Fig. 2. Isometric view of the rigid wall impact It is assumed that 50% of middle portion of the SDB wes impacted by a rigid wall having 10 kg mass and 5 ms velocity and it strikes to the SDB in Y direction as illustrated in Fig. 2. The total termination time of each test is chosen to be 25ms representing the enough time to release the kinetic energy completely.

SEA AND PL VARIATIONS FOR DIFFERENT SDB GEOMETRIES TABLE II

Fig. 5. Crash force distribution for various orientation degrees of the SDB As a result, Fig. 5 compares PL values for each angle during the impact. It can be seen obviously that the lowest PL belongs to the SDB arranged with 90 degree with respect to the horizontal line. In fact, the arrangements of @ has consequential effects on how the SDB profile resists the impact. In this scenario, if the SDB is placed horizontally (9 =90), it would create the highest value of the stiffness so the crash force would be raised also. In this example, with considering 6 degree of 90, the minimum possible of the crash force is achieved showed in the bellow graph.

Fig. 4. The angle of the impact between SDB and rigid wall In automobile accidents, the main aim is to improve safety features. In this respect, the designers have always struggled to reduce the crash pulse transferred to occupants. Hence, reducing the peak load occurred in the above simplified impact is the challenge we are looking for in this section. To examine the effects of SDB orientation angle on the created peak load (PL), three angles including 0, 45 and 90 degree toward the rigid wall impactor are assumed and demonstrated in Fig. 4.

Meanwhile, the mentioned objectives are evaluated simultaneously to find the two optimal independent geometrical factors. However, the process of the optimization is discussed specifically in the next section. Fig. 6. The cross-section of the elliptical SDB with two radii of ‘a’ and ‘b’

RSM is a method for illustrating the correlation between multiple variables as an input and output. For the specific objective functions, like SEA and PL in this study, they are assumed in terms of basis function [45], [46], [47]:

Any further improvements on SEA must sacrifice the PL and vice versa. In fact, any points in Pareto frontier can be the optimal case which is up to the designer depending on its significant level. To generate the Pareto frontier, the multi-objective optimization using Genetic [50] Algorithm of MATLAB language programming was used.

where the symbol of U and L represents the maximum and the minimum value for the functions respectively. By In this method, the cost function is constructed by relative efficiency of the two mentioned objectives:

Figs. 7. The RS (response surface) of SEA (a) and PL. (b)

The design space for the structural characteristics always consists of the link length dong with the remaining parameters depended on the type of cross section being considered. Here, hollow rectangular, hollow square and hollow circular, C-channel and I- channel cross-sections types are considered as shown in Fig. 1. The number of structural constraints is dynamic and depends on the type of cross section being considered in analysis.Physical structural constraints are also imposed during analysis, as the inner cross-sectional dimensions cannot be equal or larger than outer dimensions shown in Fig. 1. For instance, a manipulator designed with C-channel cross-section link has the additional dynamic constraints t <h and t <h/2. Fig. 1. Cross-sections considered in the analysis

When one set of design variables is analyzed, one function evaluation is completed. The calculated joint torque and manipulability measure provide the fitness evaluation for optimization. The schematic flow chart is shown in Fig. 2.

The stiffness matrices for each element are assembled to generate the global stiffness matrix. The assembled global stiffness matrix for the 6-element manipulator, K, after reduction due to constraint degrees of freedom is given by: International Review of Mechanical Engineering, Vol. 8, N. 1

Fig. 3. Forces and moments for structural analysis III. Evolutionary Techniques

Fig. 4. Schematic representation of genetic algorithm operation They combine survival of the fittest among the string structures with randomized yet structured information exchange to form a search algorithm with innovative flair of natural evolution. MOGA starts with a random creation of a population of strings and thereafter generates successive population of strings that improve over time. The processes involved in the generation of new populations mainly consist of the following operations illustrated in Fig. 4.

Fig. 5. Flow chart representing the optimization process using the proposed multi-objective genetic algorithm strategy In MOGA, its probability of occurrence is generally kept small, as a higher occurrence rate would lead to a loss of important data. MOGA, with 100% mutation rate becomes random search in the solution space. Flowchart in Fig. 5 shows the processes of the proposed MOGA to find an optimum solution.

Fig. 6. An iteration of the NSGA-II procedure TII.3. Multi-Objective Differential Evolution (MODE) MODE [2] can be categorized into a class of floating- point encoded evolutionary algorithms. The theoretical framework of MODE is very simple and MODE is computationally inexpensive in terms of memory requirements and CPU times. Thus, nowadays MODE has gained much attention and wide application in a variety of fields. MODE’s advantages are: simple structure, ease of use, speed and robust.

Fig. 7. Representation of Fuzzy Membership Function for minimization of objective function

The proposed evolutionary techniques are used in structure design of 6-link planar robot (6 dof) for pick and place operations. In the presented examples, parameters are expressed in SI system and angles in Jegrees. A 6-link planar manipulator with six revolute joints is shown in Fig. 8. The task specification and constraints are given in Table I. The constraint values for the design variables, the link lengths and cross sectional parameters of each

The variables limits used for the hollow circular cross- section are as follows: cross-section are defined to be the same for all three optimization approaches.

The results of average fuzzy membership function value (Jlavg), Solution spread measure (SSM), ratio of non-dominated individuals (RNI), optimiser overhead (OO) and algorithm effort obtained from MOGA, NSGA-II and MODE are listed in Tables VII-XT. These show the values of the optimum design variables for each cross-section, individual objective functions and the combined objective function (f). It is observed that NSGA-II approach gives smaller torques (z,); higher manipulability measure (Zz) and minimum combined objective function (f,) in majority of cases. Minimum torque requirements are for I-section for 6-link planar robots, when compared to other cross sections. 6-LINK HOLLOW RECTANGULAR CROSS - SECTION

SOLUTION SPREAD MEASURE (SSM) RESULTS OF 6-LINK PLANAR ROBOT RATIO OF NON-DOMINATED INDIVIDUALS (RNI) RESULTS OF 6-LINK PLANAR ROBOT

AVERAGE FUZZY MEMBERSHIP FUNCTION (ave) RESULTS OF 6-LINK PLANAR ROBOT

ALGORITHM EFFORT RESULTS OF 6-LINK PLANAR ROBOT From Tables II-XI, it is observed that NSGA-II technique gives minimum combined objective function (f.), maximum average fuzzy membership function (Hag), Minimum solution spread measure (SSM), maximum ratio of non-dominated individuals (RNI) and minimum algorithm effort than MOGA and MODE in majority of cases.

Fig. 9. Pareto optimal fronts obtained from MOGA, NSGA-II and MODE for 6-link planar robot (I- section)

MODE PARAMETER ANALYSIS (CR VS FS) FOR 6-LINK PLANAR ROBOT WITH I-CHANNEL CROSS SECTION generations=100. The convergences of the proposed evolutionary algorithms are within 100 generations. So the number of generations for all the algorithms is fixed as 100. The above analyses are done for the sample case Q1 =A2 =0.5. TABLE XVI

be repeated here. The evaporating liquid is FC72. The physical configuration of the problem described in the paperis shown in Fig. 1. The substrate is a silicon wafer o 0.5mm in thickness with a kapton layer of 40m on top o it This physical configuration is then translated into a numerical computation domain as shown in Figs. 2. r and z are the radial and axial direction respectively, n and t are the unit nommal and tangential vectors, 0 is the contact angle between the droplet and wall and h(r,t) is the height of the droplet that is a function of droplet radius and time, t. The computation is based on an axisymmetric two-dimensional model in cylindrical coordinate system.

International Review of Mechanical Engineering, Vol. 8, N. 1 where a is the volume fraction of liquid; u, and u, are the velocity components along the radial, r direction and axial, z direction, respectively; o is the density and rh is the mass flow rate. The subscript 1 refers to the liquid phase.

where p is the pressure, 1 is the viscosity and g is the gravitational force. The subscript g refers to the gas phase. The effect of surface tension, o, is included in the last term of Eqs. (2) and (3). The static contact angle, Ay is specified under the curvature, «, fonmulation:

The energy equation: where C, is the specific heat, T is the temperature, kis the thermal conductivity and hy, is the latent heat of vaporization. The specific heat, c, and thermal conductivity, kis given by: Ue VILIGLY UNLUSIOl! COCLIICICIIL DOLWWOCTI PU / 4 ain ali. The rate of mass and energy transfer between phases is assumed to be a function of the temperature difference:

A constant static contact angle, 6,=8° was specified [8]. The difference between the initial radius and height for experiment and the simulation is due to the number and size of the mesh used in the droplet region of the computational domain. The experiment assumed that the droplet temperature was the same as that of the ambient [8]. However, based on the infra red (IR) images of the experiment, the droplet temperature just before the droplet touches the surface of the substrate has about a 3°K temperature difference with the substrate. Hence, at t=0, the simulation assumes the droplet to be 3°K higher than the substrate and ambient.

Fig. 3. Droplet IR image shown here with the reference marker Fig. 4(a) shows one of the results of the many experiments conducted. It shows the contact line radius as well as the residual mass together in the same graph from the beginning the droplet touches the substrate until the droplet totally evaporates. Fig. 4(b) shows the enlarged portion of it showing only the assumed constant contact line region. The portion of the plot where the mass is decreasing is used for comparison with the simulation results. We concentrate on the part where the mass decreases because this simulation has been modeled to depict that the droplet evaporation has already started and hence the mass will decrease. —"s mm aa.

Figs. 5(a) Residual mass of the droplet for simulation time up until 2 seconds. (b) Mass profile forthe constant contact line region. tcc, is the total time that the constant contact line region occurs

mass of the droplet with time. Comparison between the gradients of the simulation and experimental linear plots showed about 0.58% difference. This result validates the simulation results discussed henceforth. The following discussion on the results of the simulation is based only on the constant contact line region. Figs. 4. (a) Experimental plot of the contact line radius with the residual droplet mass. tex refers the maximum radius, Mma refers to the Maximum mass, tnx is the total evaporation time. (b) The enlarged constant contact line region of Fig. 4(a)

Fig. 6. Comparison of the residual droplet mass between simulation and experiment. Two experimental results were compared to the simulation.

Fig. 9. Temperature field in and around the droplet in the constant contact line region. The time Os is taken to be the start of the constant contact line region If we relate this observation to the heat flux explained in the previous part of the paper, initially the heat flux was high because the temperature difference inside the droplet is higher but at a later time, the heat flux decreases because the temperature difference is lower.

Figs. 7. (a) Temperature along the surface of the substrate obtained from the simulation. (b) Temperature obtained directly from the top view of the IR images

Fig. 12. Concentration profiles at t=Os and t=0.8s during the constant contact line region. The concentration of FC72 vapour increases its concentration in air as time increases Fig. 11. The flow field direction at time =0s. This shows the flow movement inside and around the droplet

Fig. 10. Temperature iso-contour at time =0s and 0.8s

Nevertheless, we can still observe flow pattems inside the droplet through the side view of the droplet. A few 2-D simulations have been shown to be able to show the occurrence of hydrothermal waves inside the droplet [18], [19]. This particular simulation also observed some wave pattems believed to be hydrothermal waves inside the droplet. These waves seem to dictate the movement of the fluid inside the droplet. In Fig. 11, the flow inside the droplet seems to start at the top centre of the droplet and flows down the centre reaching the bottom of the droplet.

Tt has been theorized that evaporation flux would be highest at the droplet edge along the substrate. However, in this simulation, the evaporation flux is also high at the top centre of the droplet. The evaporation flux decreases after this point onwards. They obtained qualitatively similar concentration fields. The concentration of the droplet decreases with time. In tum, the concentration of FC72 vapour increases its concentration in air as time increases. Figs. 13 show the concentration profile of the droplet above its surface. The plots show the concentration along both the r=0 and z=0. The concentration of the FC’72 decreases with time at the droplet edge (Fig. 13(b) as well as the top centre surface of the droplet (Fig. 13(a)). The evaporation flux which was calculated based on the following equation: Figs. 14. The corresponding evaporation flux along r=0 and z=0. The peak of evaporation flux occurs along the surface of the droplet

Fig. 1. The forces distribution in engine mechanism when it is operated of the crank (element 1) The method is applied separately for two distinct situations: when the engine is working on a compressor and into the motor system. For the two separate cases, two independent formulas are obtained for the engine yield. We starting with the engine main mechanism in compressor system (when the motor mechanism is acting from the crank; see the Fig. 1). Now we are going to watch forces distribution in this case (Fig. 1).

Fig. 2. The forces distribution in engine mechanism when it is operated of the piston (element 3) Only are retained relations (7) in the calculation of mechanical efficiency for the two situations of the mechanism Otto cycle engine, when mechanism working in arrangements by the compressor, and then when mechanism working in arrangements by the motor

Fig. 3. The heat yield of Camot cycle The yield heat shield (to be used as a general tule the one given by Camot cycle) is a function of the average emperature of the engine (see diagram in Fig. 3).

Copyright © 2014 Praise Worthy Prize S.r.l. - All rights reserved It can be calculated now approximately the mechanical efficiency of engine mechanism in scheme by the motor, through the integration of instantly yield (relations system (6)):

The old classic engines yield can be improved by using a two stroke engine or a four stroke in V engine. — Consider below this value of thermal efficiency which may be obtained for thermal engines intemal combustion [24], but it's difficult, and especially to keep those with extemal combustion, will be determined the values of final yield of an intemal combustion engine for three individual cases: a two-stroke engine (Lenoir) [27], a four-stroke engine (Otto or diesel) normally [28]-[29], and one four-stroke in V (see the relationship (8)):

In the design of multi stage mixed flow submersible pump the bow-impeller axial gap plays a vital role in designing the required flow and head values. The solid modeling of the Bowl and Impeller is drawn using by Solid works 2010 software (Figs. 3 and 4). Bow is a stator which has eight stationary blades and the impeller is a rotating element which has six movable blades. The specifications of bowl and impeller in the (Tables IIT and IV). The bow- impeller axial gap of the existing model is 17 mm. The Modified (Increased and Decreased) model of six different Bowl-Impeller Axial Gap are 2 mm, 5 mmand 10 mm (Table V) and (Fig. 5). The hydraulic performance of a multi stage mixed flow submersible pump has been tested experimentally and the specification of the pump in the (Tables I and II). The running speed of the pump is 2880 rpm which has 7 h.p motor.

Fig. 5. Modified (Increased and Decreased) model of six different Bowl-Impeller Axial Gap III. Materials and Methods

Fig. 1. Cross Sectional View of Submersible Pump Assembly

The last two terms in the left hand side of Eq. (2) are the effects of the coriolis and centrifugal forces due to the rotating frame of reference.

Fig. 6. Surface Mesh of Submersible Pump Assembly Fig. 7. Contours of Static Pressure and Velocity Vectors for the Existing Model

It has been shown in the previous paper J. Manikandan et al [10] that the best practices of Computational Fluid Dynamics (CFD) for the problem wes found that k-e Realizable Model is predicted when comparing to the other turbulence models. K-e model solves two transport equations for accounting 1, into the conservation of momentum equation: where k is the turbulent kinetic energy, « is the dissipation rate, 1 is the laminar viscosity, py, is the turbulent viscosity, G, is the generation of turbulent kinetic energy due to the mean viscosity gradients, ox, and o,, are the turbulent prandtl numbers and C;, =1.44, Cy, =1.92 and C,, =0.09 are the constants of the model.

Tt was found that the geometry is complex: its surface is discretized with ti dements and volume with tetrahedron elements. To capture boundary layer prism elements are added near the walls. The maximum equiangular skewness for surface mesh is kept between 0.6 and 0.85. The Surface mesh of the Multistage submersible pump (Fig. 6) is generated in American National Space Administration (ANSA). The Contours of Static Pressure and Velocity Vectors for the Existing Model (Fig. 7). Then the volume mesh is generated in T- Grid. The volume mesh count for this case is 26 lakhs.

Water at ambient condition is to be working fluid property, and its analyzed parameters in the Table VII. VI. Results and Discussion To determine the performance curves for discharge, nead and efficiency are plotted in the (Figs. 8, 9 and 10).

This reduces the flow rate and increases the head value. This is because of minimum shear stress which is induced between bowl and the impeller (Fig. 12). Fig. 12. Contours of Shear Stress (Existing Model, 2mm, 5mm. and 10mm) Increased Model From the CFD analysis of various models it was Clearly seen that when the bow-impeller axial gap is increased, the size of the recirculation region increases (Fig. 11). The velocity and the static pressure are found to be minimal in this recirculation region and the flow moves from high pressure zone to low pressure zone. VI.2. Effect of Decreasing Recirculation Region of the Bow-Impeller Axial Gap:

VI.1. Effect of Increasing Recirculation Region of the Bow-Impeller Axial Gap:

Copyright © 2014 Praise Worthy Prize S.r.l. - All rights reserved Energy intended for pumping water is diverted and consumed by the impeller grinding itself into the pump bowl. This contact causes permanent damage to both the bowl, impeller and shortens the service life of the pump.(c) Thus finding the optimum bow-impeller axial gap is very essential for these types of pumps.

Aluminium alloy 7075 is an aluminium alloy, with zinc as the primary alloying element and less than half a percent of silicon, iron, manganese, titanium, chromium, and other metals. It is strong, with strength comparable to many steels, and has good fatigue strength and average machinability. Due to their high strength-to-density ratio, Al 7075 is often used in transport applications, including marine, automotive and aviation. Their strength and light weight is also desirable in other fields. Rock climbing equipment, bicycle components, inline-skating-frames and hang glider airframes are commonly made from Al 7075 aluminium alloy.

The hardness test was carried out in lab environment as per IS: 1586 - 2000. The Rockwell Hardness of each sample was investigated and the results are tabulated in the Table V.

Tt was found that the hardness of the aluminium and Tungsten carbide composite increased gradually as the percentage of WC mixture was increased. Whereas the Aluminium and fly ash composite’s hardness increased initially and as the Percentage of Fly ash was increased, a drop in the hardness was observed. The hardness of the Hybrid Metal Matrix Composite had a drastic drop in the hardness compared to the other MMCs,

A stir casting setup consists of a resistance Muffle Fumace and a stainless Steel stirrer assembly. The stirrer assembly was connected to a variable speed vertical drilling machine with range of 50 to 1000 rpm by means of a steel shaft. The stirrer wes made by cutting and shaping a Stainless Steel block to desired shape and size manually. Clay graphite crucible of 1.5 kg capacity was placed inside the fumace. The graphical representation of stir casting was shown in Fig. 1. Approximately 500gm of alloy in solid form (plate) was melted at 800°C in the resistance furnace. Preheating of reinforcement (Fly Ash at 450°C, tungsten carbide at 780°C) was done for one hour to remove moisture and gases from the surface of the particulates. The stirrer was then lowered vertically up to some centimetres from the bottom of the cnucible.

Fig. 4. SEM image of Sample 1 The scanning electron microscope (SEM) is a type of electron microscope that images the sample surface by scanning it with a high-energy beam of electrons in a raster scan pattem. The electrons interact with the atoms that make up the sample producing signals that contain information about the sample's surface topography, composition and other properties such as electrical conductivity. The SEM image of the fabricated composite shows the distribution of WC and Fly ash particles in uniform pattem and no obvious micro porosity defect are observed in all the selected samples, as shown in the Figs. 4, 5, 6.

However there was not much improvement in the value of tensile strength of the Hybrid metal matrix composite. The Tensile strength values of the various samples are tabulated in the Table VI.

MACHINING PROCESS PARAMETERS AND THEIR CORRESPONDING LEVELS

The work piece thickness 0.4 mm, machining current 0.64 as maintained. Table II shows the machining parameters and their level indentified for this investigation. Electrochemical micro machining (EMM) characteristics (MRR and Overcut) as outout responses

The electrolyte supply system consists of filter and pump arrangement. A pulsed power supply of 20 V and 30 A with capability for varying voltage, current, and pulse width was used [15]. The electrolyte of varying concentrations used in this study was sodium nitrate (NANos) and Al-10%wt SiC, - of thickness of 0.4 mm as work piece. Based on the literate review and preliminary experiments conducted, the initial process parameters and their corresponding levels are chosen.

EXPERIMENTAL RESULTS FOR: Ly7 ORTHOGONAL ARRAY Minitab 16 statistical software has been used for the analysis of the experimental work. The software studies the experimental data and then provides the calculated results of signal-to-noise ratio.

Fig. 3. Optical image for Micro hole 7V / 30 g/l / 50 Hz Table IV shows the response table for S/N ratio of MRR and its ranking order. Fig. 4 shows the main effects at each level. it can be seen that the optimal values for maximum MRR were machining voltage of 7 V, electrolyte concentration of 24 g/l, and frequency of 50 Hz. The MRR increases with an increase in pulse frequency then the dissolution efficiency increases rapidly, causing a rapid increment of MRR in the machining zone. Fig. 5 shows the residual plot for MRR.

TAGUCHI ANALYSIS: OVER CUT VERSUS VOLTAGE, ELECTROLYTE CONCENTRATION, FREQUENCY RESPONSE TABLE FOR SIGNAL TO NOISE RATIOS (SMALLER IS BETTER)

Fig. 4. S/N ratio graph for MRR Fig. 5. Residual graph for MRR

TAGUCHI ANALYSIS: MRR VERSUS VOLTAGE, El ELECTROLYTE CONCENTRATION, FREQUENCY RESPONSE TABLE FOR SIGNAL TO NOISE RATIOS (LARGER IS BETTER)

S = 0.0126383 R-Sq = 96.31% R-Sq(adj) = 95.20%

ANOVA TABLE FOR OVER CUT S = 20.7233 R-Sq = 84.64% R-Sq(adj) = 80.03%

Fig. 8. Interation Plot for S/N Ratios of MRR Fig. 9. Interation Plot for S/N Ratios of Overcut

Fig. 1. Semi-finished product of coconut coir fibre pressed mat The functions of these plates and frame are to compress the fibre after epoxy is applied, maintain specimen thickness, and also as a cover to avoid the debris from entering into composite parts during the curing time. The mould then was cleaned and release agent (wax) was applied on the mould before being lay- up with the fibre.

Fig. 2. Random oriented pressed mat coir filbre/epoxy composite Fig. 2 shows the sample of fabricated panel. Six categories of samples were prepared for different fibre weight ratio ranging from 20%, 25%, 30%, 35%, 40% and 50% with epoxy resins.

For the tensile test, the. samples which dimension of 250 mm x 25 mm x 2.5 mm were prepared and test under room temperature and test speed of 2 mm/min on a 10 KN Instron Universal Testing Machine to obtain average Young Modulus for each fiber weight percentage. The three point flexural test was done by using 5kN Instron Universal Testing Machine test speed of 2 mm/min under room temperature. Figs. 3. Mechanical tests: (a) tensile test (b) flexural test (c) torsion test

Therefore, the shear modulus data obtained could be used for further supporting the tensile modulus and flexural modulus data obtained which agreed that the higher than 35% coir fiber loading panels shown the higher flexibility behaviour which are soft and deformable while the less than 35% wt coir fiber loading panels show more rigid behaviour which are stiff and relatively hard. Fig. 6. Variation of the average shear modulus with the mass fraction of coir fibre

The results than compared to the flexural strength for pressed mat coir fiber-polyester composite under two different fabrication compression pressures which are 2.6 MPa and 5.2 MPa that was done by Monteiro et. al [9] as shown in Fig. 5. This suggests the strength tends to decrease with the amount of fiber and reveals that the randomly oriented coir fibers are not reinforcing neither epoxy matrix nor polyester matrix at all. Several measurement points (25% and 35% wt fibre) had not been included in Monteiro work. Besides that, the different in pressures did not be a major influence to the value of flexural strength compared to the difference type of matrix that has been used. Fig. 5. Variation of the flexural strength with the mass fraction of coir fibre for different compression fabrication pressures

In addition, the examination of these three-phase flows can be considered as a necessary step in order to prepare simulations of three-phase natural gas-cnude oil- sea water flows that may occur within DIFIS system due to the presence of natural gas dissolved in the leaking oil from. a maritime accident.

The turbulence kinetic energy, k, and its dissipation rate, ¢, are obtained from the following transport equations: International Review of Mechanical Engineering, Vol. 8, N. 1

where p, is the density of the p® fluid. Also m,, is the mass transfer from phase p to phase q and m, is the mass For the pt phase, ‘the continuity equation for the volume fraction has the following form: In computational fluid dynamics, the Volume of fluid method is one of the most well known methods for volume tracking and locating the free surface. The motion of all phases is modelled by solving a single set of transport equations with appropriate jump boundary conditions at the interface. It can model two or more immiscible fluids by solving a single set of momentum equations and tracking the volume fraction of each of the fluids throughout the domain.

Fig. 3. Computed distributions of water (light) and air-oil (dark) volume fraction at different time steps for the first case (0.1 m/s airinlet velocity) Figs. 3 and 4 show the results of the simulations at different time steps. For better understanding, symmetry is applied to the pictures although this is an axisymmetric model. The leaking oil is initially collected in the 6m high dome.

Fig. 4. Computed distributions of water (light) and air-oil (dark) volume fraction at different time steps for the second case (1 m/s airinle velocity)

Fig. 8. Flow pattem map for vertical upward three-phase flow (0.1 mys air inlet velocity)

Fig. 7. Radial mixture velocity for different flow pattems for the second case (1 m/s air inlet velocity)

Fig. 5 shows the investigated cellular cylinder for the strength with fixations and loadings. Since the construction is vertically symmetrical, thus only a tenth of geometry is being used for the simplification of the calculations - that is, construction with one celular element. Fig. 1. Schematic image of investigation object

Fig. 3. Cellular cylindrical cross-section with the V-shaped core Fig. 2. Cellular cylindrical cross-section with comugated core

Fig. 4. Cellular cylindrical cross-section with the X-shaped core

TABLE OF CELLULAR CYLINDER DIMENSIONS FOR RESEARCH CASES TABLE I

Fig. 7. Distribution of cellular cylinder with a commgated core of reduced Von Mises stress at the case of tv=tk=ti=2 " Varying the length of the cylinder diameter, the maximum pressure gradually decreases, but the pressure, varying in the length at each 1000 mm reduces very little.

Fig. 6. Sample fixation and loading investigating stability Fig. 6 shows a researched cellular cylinder for the stability with fixations and loadings. Construction fixations remain the same as in the investigation of strength. In this case differs the loading - vacuum.

Fig. 8. Variation impact of cellular cylinder element thicknesses ty, tk, ti on the maximum sustainable capability

Fig. 14. Impact of cellular cylinder element thicknesses tv, tk and ti on the maximum sustainable capability Fig. 13. Distribution of cellular cylinder with V-shaped core of reduced Von Mises stress at the case tv=tk=ti=2

Fig. 12. Variation impact of cylinder length L on the maxinum sustainable capability Core element thickness tk, and the thickness of the outer cylinder ti have a minimal effect on the maximum sustainable capability compared with the tv impact.The increase of the thickness of the outer cylinder even reduces the maximum sustainable pressure.

Fig. 10. Variation impact of core edement b on the maximum sustainable capability After the comparison of masses of monolithic and cellular cylinders at the same maximum sustainable capability, it was revealed that a monolithic cylinder is 3,47 times lighter than the cellular one.

Fig. 15. Variation impact of core element width La on maximum sustainable capahility sustainable capability. Meanwhile, the increase in the width of the core member the maximum sustainable capability increases and after increasing it to the width when core members support each other - sustainable capability reaches the maximum value.

As can be seen from the graphs, provided in Fig. 20, tv thickness of inner cylinder has the greatest impact on the maximum sustainable capability. Core element thickness tk, and the thickness of the outer cylinder ti have a minimal effect on the maximum sustainable capability compared with the tv impact. Fig. 19. Distribution of cellular cylinder with X-shaped core of reduced Von Mises stress at the case tv=tk=ti=2

Fig. 20. Impact of cellular cylinder element thicknesses tv, tk and ti on the maximum sustainable capability

Fig. 16. Variation impact of core element b on the maximum sustainable capahility

Fig. 18. Variation impact of cylinder length L on the maximum sustainable capability Fig. 17. Variation impact of inner cylinder D; on the maximum sustainable capahility

Fig. 25. Image of buckling of cellular cylinder with a corrugated core at case tv=tk=ti=2

Fig. 23. Variation impact of inner cylinder D, on the maximum. sustainable capahility Fig. 24. Variation impact of cylinder length L on the maximum sustainable capahility

Fig. 26. Impact of cellular cylinder element thicknesses ty, tk, ti on the buckling

Fig. 22. Variation impact of core dement b on the maximum sustainable capahility

Fig. 21. Variation impact of core element width La on the maximum sustainable capability Varying the diameter of the cylinder, sustainable pressure decreases gradually. Varying the length of the cylinder diameter, the maximum pressure decreases, but this decrease is very little.

Fig. 27. Variation impact of core element width La on maximum sustainable capahility

Tt can be concluded from the obtained graph that the rise of the core element increases the maximum sustainable pressure. Varying the diameter of the cylinder, sustainable pressure decreases gradually. Fig. 30. Variation impact of cylinder length L on the maximum sustainable capability

Tt can be concluded from the obtained graph that the rise of the core element increases the maximum sustainable pressure. After increasing core height b to 95mm the extemal pressure reaches a maximum Varying the diameter of the cylinder, maximum pressure decreases gradually. Fig. 31. Image of buckling of cellular cylinder with a V-shaped core at case tv=tk=ti=2

Fig. 32. Impact of cellular cylinder element thicknesses tv, tk the on the buckling

Fig. 28. Variation impact of core dement b on the maximum sustainable capahility

Fig. 29. Variation impact of inner cylinder D, on the maximum sustainable capahility

Fig. 38. Impact of cellular cylinder element thicknesses tv, tk, ti on buckling Fig. 37. Image of buckling of cellular cylinder with X-shaped core at the case tv=tk=ti=2

Fig. 39. Variation impact of core element width La on maximum sustainable capability

Fig. 34. Variation impact of core dement b on the maximum sustainable capability

Fig. 36. Variation impact of cylinder length L on the maximum sustainable capability

cellular cylinders at the same maximum sustainable capability, it was revealed that a monolithic cylinder is 2, 15 times lighter than the cellular one. Case of mass ratios was investigated, when cellular cylinder dimensions are of case tv = tk = ti = 2. After optimization of the cellular cylinder for minimum weight, the monolithic cylinder has become 2,14 times lighter.

Fig. 35. Variation impact of inner cylinder D; on the maximum sustainable capahility

Varying the diameter of the cylinder maximum pressure decreases gradually. Fig. 42. Variation impact of cylinder length L on the maximum sustainable capahility

Fig. 40. Variation impact of core element b on the maximum sustainable capahility In distributing the core a maximum pressure increase was observed. By increasing the width of the core from 137 to 155 mm the maximum extemal pressure reaches its peak, but the increment is not as significant as it was in the preceding steps.

Fig. 41. Variation impact of inner cylinder D; on the maximum sustainable capability It can be concluded from the obtained graph that the rise of the core element raising increases the maximum sustainable pressure. By increasing the height of the core from 85 to 95 mm the sustainable capability reaches its peak, but the increment as well as the distribution is less significant.

Copyright © 2014 Praise Worthy Prize S.r.]. - All rights reserved

In this application, we consider a turbulent flow, for two steady-state flow values Vo =0.3 m/s and Vo =1.4 mys. The flow is in the horizontal copper pipe anchored to the upstream to a tank filled with water and of height Hy, ending at the downstream to a valve that closes abruptly. The parameters of the fluid and the pipe are summarized in Table I. In this study, we analysis the evolution over time, of the pressure head, the volume of the discrete vapour cavity and friction stress.

Fig. 3. Head at the valve for Vo =0.3 m/s Figs. 1 and 2 show at the valve, for permanent flow regimes of velocity Vo =0.3 m/s and Vo = 1.4 mjs, the change in friction stress with taking into account or not the unsteady part, over time.

The boundary conditions of the problem are a pressure head imposed by the upstream reservoir height, a fast closure of the valve at the downstream end of the pipe, and the liquid-vapour pocket interface conditions H=H,,,; resulting in case of cavitation.

Fig. 5. Evolution of the vapour pocket at the valve for Vo =0.3 m/s Fig. 5 shows, in the same conditions, the behaviour of the volume of the discrete vapour cavity formed at the valve, with taking into account or not the part of the unsteady friction term.

Fig. 4. Head at the pipe midpoint for Vo =0.3 m/s

Fig. 6. Head at the valve for Vo =1.4m/s Copyright © 2014 Praise Worthy Prize Sr. - All rights reserved

Figs. 8 and 9 correspond to the evolution of air pockets formed over time at the valve and at the pipe midpoint with taking into account or not the unsteady friction tem.

This study demonstrates, clearly, the fact that the

Fig. 2. Wear test ng Taguchi method has been used to study the effect of three wear parameters (Temperature, load, sliding velocity) on wear of AISI 202. To find the optimal combinations the following step by step procedure is followed for the DOE [12]:

The disc is prepared to a diameter of 180 mm and thickness 10mm. The surface is finely finished in a surface-grinding machine. The prepared specimen is shown in the Fig. 1. Fig. 1. Real area of contact of the prepared specimen The specimen of AISI 202 grade had been purchased in the form of rod of diameter 20mm and it is reduced to 10 mm as per the mechanics of contact. The real area of contact is designed to be a diameter of 4 mm and 2mm to a length of 7mm from one end of the diameter reduced test pin. Wear tests were conducted on the pin on disc as shown in Figure 2. The pin and disc are of similar material of grade AISI202. The wear tests were carmied out at various loads (13 MPa, 18 MPa, 22 MPa and 27 MPa), sliding velocities (0.2 m/sec and 2 m/s) and at temperatures (200°C, 400°C, 600°C) and vacuum condition (600mm of Hg). The wear was measured by the loss in weight.

At very low speed, the u is quite high. Again at very high sliding speed, surface melting occurs which works as lubricant and produces a very low coefficient of friction.

L9 (3°) ORTHOGONAL ARRAY, EXPERIMENT RESULTS AND S/N RATIO TABLE V The weight loss values measured from the experiments and their corresponding S/N ratio values are listed in Table V. Regardless of the category of the performance characteristics, a smalla S/N value corresponds to a better performance. Therefore, the optimal level of the machining parameters is the level with the smallest S/N value. To obtain optimal testing parameters, the-lower-the-better quality characteristic for weight loss was taken. The S/N ratio for each level of testing parameters was computed based on the G/N analysis. Moreover, a statistical analysis of variance (ANOVA) was performed to see which test are statistically significant. With S/N ratio and ANOVA analyses, the optimal combination of the testing parameters could be predicted for a 95% confidence level. In the response Table VI, the control factor with the strongest influence was determined by difference values. The higher the difference, the more influential was the control factor. The strongest influence was found by sliding velocity(C), Temperature (A) and Load (B) respectively. The main effects and their interaction plots for S/N ratios are shown in Fig. 9. Optimal testing conditions of these control factors can be easily determined from this graph.

Fig. 6. Friction-time curve at 200°C. Comparison was made between 0.2m/s and 2m/s

Fig. 8. Friction-time curve at 600°C. Comparison was made between 0.2m/s and 2m/s

Fig. 7. Friction-time curve at 400°C Comparison was made between 0.2m/s and 2n/s

Fig. 12 reveals that the pin surface shows the mechanism of the de-lamination when the pin specimen is tested at 400°C, Vacuum condition. The wom surface consists of debris, cracks and pores.

Fig. 13. Specimen tested at 400°C These ploughs are parallel to the sliding direction. From the fractured surface analysis the wear mechanism associated with this material is plastic deformation. At high temperature, there is increase in area, throwing-out and squeezing-out of some surface films of AISI 202 in pin. Also at higher temperatures, surface melting occurs which works as lubricant.

Fig. 1. Experimental setup of Three Stage Auto Refrigerating Cascade (3 stage ARC) System

Three stage Auto Refrigerating cascade system was fabricated as per the design parameters. The detailed photographic view of the system is shown in Fig. 1 and the line diagram of the setup is shown in Fig. 2.

7 Fig. 3 shows the variation of Exergic Efficiency of the ARC system with the variation of mass fraction of R1270 and R14 in the Zeotropic Mixture of R1270, R170 and R14. The maximum value of Exergic efficiency from the Fig. 3 was 66.9% which correspond to the mixture of R1270, R170 and R14 with the mass fraction of 0.265:0.18:0.555 and the lowest among the values obtained is 55.2% which correspond to the mixture of mass fraction of 0.325:0.18;:0.495. The values of COP, Exergy Lost, Exergic Efficiency and Efficiency defect readings in individual components were calculated using the readings obtained by the experimental setup. These Parameters were discussed in this section for better understanding of the 3 stage ARC System.

Fig. 4. Variation of COP for Different Mass Fraction of R1270& R14 in the Zeotropic Mixture of R1270, R170 & R14 Fig. 4 shows the Variation in COP with the variation of mass fraction of R1270 and R14 in the Zeotropic Mixture of R1270, R170 and R14.

Fig. 9 shows the variation of Efficiency defect in each component of the ARC system with the variation of mass fraction of R1270 and R14 in the Zeotropic Mixture of R1270, R170 and R14. Fig. 9. Variation of Efficiency Defect in Each Component of ARC System for Different Mass Fraction of R1270 & R14 in the Zeotropic Mixture of R1270, R170 & R14

Fig. 8 show the vanation of Exergy lost at compressor with the variation of mass fraction of R1270 and R14 in the Zeotropic Mixture of R1270, R170 and R14. The maximum and minimum values of Exergy lost at compressor are 199.51 W and 167.46 W corresponds to the mixture of mass fraction 0.265:0.18:0.555 having the value of 270.42W of compressor work input and 0.325:0.18:0.495 having the value of 246.95 W of compressor work input.

Since the same phenomenon of exergy destroyed as per the Fig. 10 can be utilized to calculate the efficiency defect of each and every component of the ARC system represented by Fig. 9, the interpretations holds good for the effective and efficient working of Three stage Auto Refrigerating Cascade (3 stage ARC) System. Exergic Analysis and performance analysis of Three Stage Auto Refrigerating Cascade (3 stage ARC) system wes conducted on the setup operating between the evaporation and condensation temperatures of 183K ca and 301K (30°C) or over the temperature range of 190°C: —— sg , weoum ¢ ia am a i Sa a

Fig. 2. Positions of outlet strands for all types of multi-strand tundishes considered

Fig. 1. Geometry of the multi-strand tundish Figs. 1 and 2 show the positions of outlet strands and their locations with respect to the inlet gate, respectively. The outlet strand farthest from the inlet gate is termed as the F.O. (far outlet) and similar terminology was chosen for the other two outlet strands, i.e., M.O. (middle outlet) and N.O. (near outlet).

Fig. 3. Methodology adopted to compare and optimize different multi-strand tundishes This would cause change in dimensions or transient parameters for all the processes or units before the tundish. This would cause change in dimensions or transient parameters for all the processes or units before the tundish. The dimensions of all the tundishes were taken to be the same and thus the flow rate of liquid steel was maintained by controlling velocities of liquid steel at the inlet gate and outlet strand gates. The values of the properties of liquid steel used were [18]-[22]: specific density = 7010 kg/m’, specific heat = 821 J/kgK, thermal conductivity = 30.5 W/mK and viscosity = 0.007 kg/ms. An accurate representation of fluid turbulence is a challenging task in the simulation of the flow in a steel tundish. We have used the standard k-e model [23], a two equation turbulence model. In the ke model, two transport equations, one for the turbulent kinetic energy (kK) and other for its dissipation rate (c) are used to calculate the eddy viscosity. It is a widely used turbulence model for industrial applications.

The values of the constants used in the ke model are C; = 1.44, C, = 1.92, o, = 1.0, o. =1.3 and C, = 0.09 [23], [24]. The mentioned Eqs. (1) to (5) wer solved numerically using the finite volume method to obtain an Eulerian flow field. The boundary conditions described in Section III.2 were applied to obtain the flow field.

Fig. 4. A comparison of the present mixing time characteristics for the Far outlet (a), middle outlet (b) and near outlet (c) with those reported by Merder et al. [18]

Fig. 5. A comparison of the present inclusion removal with results reported by Mikki and Thomas [1]

EFFECT OF OUTLET STRAND POSITION ONTHEI LIQUID STEEL TEMPERATURE OF TUNDISH FOR THE SAME FLOW RATE OF LIQUID STEEL AT THE INLET

EFFECT OF OUTLET STRAND POSITION (ON THE INCLUSION REMOVAL EFFICIENCY OF TUNDISH FOR THE SAME MASS FLOW AT THE INLET

Fig. 6. Velocity vectors drawn on plane A for two-strand tundish having outlet at F.0. Fig. 7. Velocity vectors drawn on plane A for two-strand tundish having outlet at N.O.

The mass flow rate at the outlet strand gate of the six- strand tundish was chosen as the reference flow rate for calculating the mass flow rate at the inlet of different mullti-strand tundishes.

EFFECT OF OUTLET STRAND POSITION ONTHE INCLUSION REMOVAL EFFICIENCY OF THE TUNDISE FOR THE SAME MASS FLOW AT OUTLET STRAND TABLE IV ffADitivV EFFECT OF OUTLET STRAND POSITION ON THE LIQUID STEEL TEMPERATURE OF THE TUNDISH FOR THE SAME FLOW RATE OF LIQUID STEEL AT THE OUTLET TABLE III

The temperature drop inside the tumdish wes Maximum in the two-strand tundish and minimum in the ix-strand tundish (Table IV). aig Se te

efficiency, outlet strand temperatures, minimum band width of inclusion removal, and temperature of each strand, we can conclude that for the same capacity tundishes, a change in the number of outlet strands does not have any significant effect on the operating parameters of the tundish provided an optimum position of the outlet strand is used. For tundishes having the same mass flow rate at each outlet strands (different capacity) two-strand tundish is found to be somewhat supenior in the removal of non-metallic inclusions but temperature loss in two-strand tundish is quite high as compared to that in a six-strand tundish. Fig. 8. Comparison of mixing time characteristics for optimized two- strand F.O. tundish with same capacity (case a) and different capacity (case b)

Fig. 1. Maximum weld travel velocity, heat source spot size, and interaction time with respect to the intensity of the heat source [3] Automatic aluminum welding can be achieved by using different processes like tungsten inert gas (TIG) welding, metal inert gas (MIG) welding, and other common welding technologies. Welding processes moving at a high speed with a too short dwell time (less than 0.3 s), which varies with the intensity of the heat source, must be automated and cannot be controlled manually as shown in Fig. 1.

Fig. 2. Comparison of single wire and tandem wire MIG welding of an aluminum fuel tank [18] Aluminum is one of the most difficult materials to melt with laser [33] due to the poor coupling (absorption of the beam energy by the metal being welded), high thermal conductivity, high reflectivity, and low boiling point [4]. As the wavelength of the laser increases, the coupling becomes poorer [34]. It was reported that a Nd:YAG (solid state laser) with a characteristic wavelength of 1.06 jim provides better coupling with aluminum than a CQO, laser with a characteristic wavelength of 10.6 jim [35].

Figs. 3. (a). Schematic diagram of plasma-MIG weld system [31], [32], (b) General layout of a robotic plasma-MIG welds system [31]

+++ Excellent, ++ Very good, ++ Good, + Ok, - : Difficult, - - : More difficult ALUMINUM WELDING PROCESSES COMPARISON

Figs. 4 (a) Schema of laser-MIG hybrid system [42] (b) KUKA KS Hybridtec (Laser-MI system) [43

REAL CASE PARAMETERS OF PROCESSES FOR AUTOMATED WELDING OF ALUMINUM

Fig. 2. Membership function of inputs- Enor, Rate of Change of Error and Output Actuator Force ~The mule base is in the form of linguistic variables using fuzzy conditional statements.

The term a is the weight [17]. In traditional GM(1,1) model the value of o in B matrix or data matrix is chosen as 0.5 [16] and:

Fig. 3. Structure of Grey Fuzzy Logic Controller based VSS The future error and rate of change of enor is calculated from the predicted value and given as the input to the TFLC. The membership functions and nile table of GFLC are similar to that of TFLC discussed in section 2. The block diagram of GFLC based VSS is shown in Fig. 3. The performance of TFLC is improved by combining grey prediction algorithm with TFLC. Based on the past and present sprung mass displacement values (five sets of data), GM (1, 1) model predicts the future value of sprung mass displacement, i.e. y(k-+1).

Fig. 5. Structure of Modified Grey Fuzzy Logic Controller based VSS

Fig. 4. Flowchart of PSO algorithm The GFLC strategy only treats next position output error of the sprung mass displacement and the rate of change of the sprung mass displacement and the rate of change of error. This cannot effectively overcome the dynamic effect of the tyre deflection in the suspension system. The dynamic effect causes a bouncing phenomenon in tyre that not only decreases road-holding of the vehicle but also increases the acceleration of the vehicle body [20]. The tyre deflection always affects the road holding of the vehicle, especially when the tyre is rotating. Therefore ride comfort as well as road-holding ability of the vehicle is affected. Hence two FLCs (FLC1 and FLC2) are designed to improve both road-holding ability and the ride comfort of the vehicle simultaneously. This Modified FLC strategy involving two GFLCs is known as Modified GFLC.

The mathematical model of the quarter car defined by (1) with the parameters listed is simulated using MATLAB-SIMULINK. Dual Bump road profile is utilized for this simulation. Mathematical representation of dual bump input, shown in Fig. 6, with amplitude of 10 cm and 5 cmis given by: Simulations are conducted for open loop passive, TFLC, GFLC, Modified FLC and Modified GFLC based VSS. The prediction error of GM(1,1) model is minimized by changing the initial condition and optimizing the value of weight factor used in the computation of data matrix.

Fig. 9. Grey predicted sprung mass displacement

The parameters of PSO used for optimization of a are given in Table III. The convergence plots are shown in Figs. 7 and 8. The predicted output for spring mass displacement and tyre deflection is shown in Figs. 9 and 10.

Figs. 11-14 show the suspension parameters for open loop passive TFLC, GFLC, Modified FLC (enhancement of ride comfort and road holding ability) and Modified GFLC. Fig. 11 illustrates that the sprung mass displacement of Modified GFLC better when compared to passive and other control strategies. It is reduced to the order of 1.6 cm for 10 cm bump.

Fig. 8. Convergence plot a - GM, (1,1) model Fig. 7. Convergence plot a - GM (1, 1) model

for a signal s(t). In the evaluation of vehicle ride quality, the PSD of the sprung mass acceleration as a function of frequency is of prime interest and is shown in Fig. 15. Under the perturbed conditions, the spring mass is assumed to be increased by 30% from the nominal value, while both damping coefficient and spring rate are considered to be decreased by 30% from their nominal values.

The tyre deflection of VSS for dual bump input is presented in Fig. 14. The tyre deflection of TFLC and GFLC is increased when compared to passive system where as the tyre deflection of Modified FLC and Modified GFLC is decreased when compared to FLC and other contol strategies. In order to investigate the robustness of the proposed controller, the perturbations of the system parameters have been included.

Fig. 15. Power Spectral Density of Sprung Mass Acceleration Both GFLC and Modified GFLC have significantly suppressed the acceleration of sprung mass effectively in the low frequency band. It can be observed from the PSD plot that the sprung mass acceleration has been brought down within the frequency range of 0.4 Hz to 8 Hz for dual bump input. Both GFLC and Modified GFLC perform much better than passive and TFLC. Modified GFLC exhibits superior performance especially in the human sensitive frequency range of 4 Hz to 8 Hz.

Instrumentation, Velammal Engineering College, Chennai. Her areas of interest are intelligent controllers, and mechatronics. She is a life member of ISTE.

‘Then the air reaches the level of the bypass (13) and can go to either in or outside the dryer. It is directed outward when the weighings are carried out. The rest of time, it is sent on a sieve (14) in order to promote a patch then on a honeycomb (15) that will propel the air along a vertical axis. These two systems will cause artificial loading loss and homogeneous plug flow for performing identical drying, in height and diameter. In a period of time which is fixed, weighings are performed: the support of carrots (drying rack) (17) will descend and rest on a system (tripod) connected to the balance (21) then it goes back. The sensors are positioned at different locations ((7)-(12), (16), (18) and (19)) on the system and connected to a computer that will provide instructions and collect data and the system can be operated automatically or manually. The electrical signals are so high (several Amperes), and are attenuated by static relays, so that the computer collects without being damaged. Safety systems are present: the thermostat (7) will cut operating if the temperature in the pipe is too high, and regulators in boxes to minimize the strong electrical signals and adjust temperature to desired values. The principle of a thermocouple ((8), (10), (16) and (18)) consists of the measurement of the electromotive force (in micro volts) between two different materials wires, this measure is proportional to the temperature and the length of wires, hence the need for calibration. In food processing, we advocate more the use of PI100 sensor (11), where its principle is Measurement of resistance that changes with temperature.

Fig. 3(a). Evolution of water content with time After 30 minutes of drying, for example, X =2 [kg water/ kg DM] at 50°c and X =1 [kg water’ kg DM] at 70°c. Comparing changes in water content over time X, for temperature drying of 70°C and 50°C, Figure 3(a) shows a decrease in water content over time and when the drying is done for 50°C, it is slower than 70°C.

From Fig. 2, it is clear that the product mass loss is very important and faster for carrots.

Fig. 3(b). Comparative water content variations of carrots and potatoes over time

Fig. 5. Evolution of the drying rate as a function of the water content

Fig. 7. Evolution of relative humidity of the air over time

Fig. 6. Evolution of the air temperatures with time

Fig. 4. Evolution of the drying rate as a function of time Here potatoes retain more water than carrots, so it is more difficult to remove water the bonds and vice versa, it will be easier to rehydrate. Curves for the drying rate as a function of time t: the curves of Fig. 4 are similar for the two temperatures, the drying rate decreases with time. Indeed, initially the product contains a lot of water so the speed is high.

Fig. 8. Evolution of the drying rate at different air velocities as a function of time At the same time, there will be less water to remove from carrots. By using other data with sliced carrots at T=60°C but at speed air of 1.5 and 2.5m/s, similar kinetics were obtained: the air velocity has no significant influence on the washers drying (Fig. 8).

Also, for data with grated and washers carrots at 70°C and 1 m/s (Fig. 9) led to the conclusion that the grated camot drying is much faster and this is due to the increased surface heat exchange when using grater camots. Grated carrots are dry in 25 minutes and sliced ones in 50 minutes.

Fig. 2. Normal probability plot residuals for O.C.

Fig. 3. Influence of applied voltage and flow rate on MRR

Fig. 4. Influence of applied voltage and tool feed rate on MRR

ANALYSIS OF VARIANCE For O.C Fig. 1. Normal probability plot residuals for MRR

Fig. 5. Influence of applied voltage and flow rate on O.C

Fig. 6. SEM Micrograph of machined hole at 18 volts

Fig. 7. Influence of applied voltage and tool feed rate on O.C

Fig. 8. SEM Micrograph of machined hole at 0.6mm/min

Hard turning experiments were performed in KIROLOSKAR model conventional machine. A chuck rotated in different spindle speed and a tailstock in order to support the work piece was placed with the center lathe. In this experiment, on the one hand mechanism rotates work piece around its own axis, while the table of the center lathe is feeding to the rotating cutting tool linearly. The cutting tool is rotated itself axis around. The cutting tool and work piece contact each other in the tangential position. In the experimental studies, AISI M2 Steel having dimensions as 40 mm diameter and a length of 350 mm were used. Figure 1 shows the experimental set up (Kiroloskar model machine). Fig. 1.Experimental setup (Kiroloskar Model Lathe)

DESIGN OF EXPERIMENT MATRIX AND MACHINING CHARACTERISTICS TABLE III

The normal probability plot is presented in Figure 2. It is noticeable that residuals fall on a straight line, showing that the errors are dispersed and that the regression model completely matches the observed values. To validate the above surface roughness models, the predicted values have been plotted with the experimental values for different combination of machining parameters as shown in Fig. 3. The straight line shows the ideal trend and dots represent the observed values. It has been found that the predicted surface roughness is very close to the observed values and hence the results obtained by the regression

Std Dev= 0.020914, R-Square-0.999919, Mean=11.521405, Adjusted R-Square= 0.999845, Percentage of C.V= 0.181522, Predicted R- square- 0.999567, PRESS- 0.023268, Adeq. Precision- 301.2916

Among many combinations of process parameters, the optimum combination can be selected from this contour graph. Based on the statistically analysis above, it can be drawn that the relation between Tool life and the variables tested cutting speed, feed and depth of cut can be modeled mathematically. The equation (10) that depict these relations as follows:

Ye. LULIULLULAY LULGO Fuzzy logic which is recognizing and identifying systems, has been developed and widely used have shown many practical applications of fuzzy logic systems in machine monitoring and diagnostic. This approach is more spontaneous and does not need complicated mathematics. With this technique many fields are gaining benefits. Fuzzy rules employ fuzzy set theory which is used to develop fuzzy rule based model. Generally construction of rule base is done by using two types of fuzzy logic rules such as Mamdani type or TSK type. Emergent competitive market trend forces the industrial sectors to undergo importantly for process optimization of all the processes is not free from fretfulness, because of the much complexity involved in many processes .The machining parameters obtained from standard manuals or handbooks resulting from experiments conducted under certain environmental conditions cannot be directly used for few alloys materials. In this experiment, the selection of appropriate machining parameter for the minimum _ surface

A fuzzy logic tool box graphical user interface available in MATLAB was used for training the experimental values in the research work. The kind of the input membership function is Triangular MF form as follows. In this fuzzy logic Mamdani system one member ship function is related to each rule. Fuzzy rules in the system based on the assumption that according to the reduction in feeding rates and depth of cuts and also increasing in cutting speed, then surface finishing will be improved. For example, a fuzzy rule is considered as follows. If (v is high)and(f is low) and(d is low) then(Ra) is L1) . Similarly for tool life If (v is low) and (f is low) and (ap is low) then T, is H1). The structure of fuzzy logic modeling with three input parameters and one output parameter is shown in Fig. 22. Comparison of surface roughness (Ra) and Tool life(TL) for RSM and Fuzzy logic with Experimental result is shown in Tables VI and VII.

COMPARISON OF SURFACE ROUGHNESS (Ra) FoR RSM AND FUZZY LOGIC WITH EXPERIMENTAL RESULT

Fig. 23. Comparision between Experimental, RSM and Fuzzy logic value for Surface roughness Fig. 24. Comparision between Experimental, RSM and Fuzzy logic value for Tool life

to return a set of optimal solution. MATLAB is a high performance language for technical computing, which is used to optimize the objective function of AISI M2 steel.

The minimization of the fitness functions value of(14)and and maximization of the fitness functions value(18)are subjected to the boundaries of the boundaries (Limitations) of the process parameters. Control parameters and their level in Table II is selected in present the limitations of the optimization solutions and is given as follows. Basically to obtain the optimal solutions some criteria must be considered by genetic algorithms (GA) as listed in Table VIII. By using the fitness functions formulated in (14) and (18), the limitations of process parameter formulated in (15), (16) and (17) and (19), (20) and (21) the genetic algorithm parameters given in Table VIII, the MATLAB optimization tool box is next applied to find the minimum values of (Ra) and maximum value of (TL) at the optimal points. The results of the MATLAB optimization tool box are given in Figures 25 and 26 for surface roughness and Figs. 27 and 28 for Tool life.

From Fig. 25, the set values of optimal process parameters that lead to minimum (Ra) value are 120.2951 m/min for cutting speed, 0.04856 mm/rev for feed 0.4423mm for depth of cut. Figure 26 shows the best fitness value of the GA is 0.604014 micrometer with the mean fitness value being 0.604014 micrometer. Fig. 26. Fitness functions plot of GA for surface roughness From Fig. 27, the set values of optimal process parameters that lead to maximum tool life (TL) value are 90.3507m/min for cutting speed 0.1406389 mm/rev for feed 0.1248070 mm for depth of cut. Figure 28 shows the best fitness value of the GA is 13.2092 minutes with the mean fitness value being 13.2092 minutes.

In this investigation, the wear and friction behavior of commercially used Cast iron and the MMC disc and their counter face commercial automobile friction material were studied and compared. The inner diameter, outer diameter and the thickness of the discs were 155 mm, 230 mm and 7 mm respectively. The surfaces of the discs were machined to an average roughness value of 1.5m which is same as the roughness value of the sliding surface of the actual commercial brake rotor. The composition of the C.I material is shown in Table I. The composition of the aluminium alloy is shown in Table IT. International Review of Mechanical Engineering, Vol. 8, N. 1

At minimum loads the contact pressure and temperature rise are low, so that at lower loads the minimum wear rate is observed. Higher wear is observed for the maximum load. The same trend is observed for the increased sliding velocity. Fig. 6. Variation of wear rate of cast iron disc with applied load

Fig. 4. Experimental set up used for wear test Tests were conducted at different loads in the range of 30-110N and at a linear speed range of 3-9m/s for 30 minutes. The friction coefficient was calculated from the friction force measured during the wear test. During each test, the contact temperature was measured, by using chromel-alumel thermocouple inserted in the friction material pin at a distance of 2mm from the contact surface. The discs and pins were weighed before and after each test by using a microbalance having an accuracy of 0.1mg and the weight loss was used to detenmine the wear rate.

“From Fig. 5 it can be observed that the friction material consists of number of elements viz. C, Fe and Cu and compounds like CaCo3, TiO and ZnO. III. Results and Discussion

A commercial automotive brake material was used as pins for the wear test. The pins were machined such that each pin was of 10mm diameter and a height of 25mm as shown in Fig. 3. The pin is fixed on a pin holder, for mounting it on the machine. The surface is polished by using A320 emery paper. The surface is cleaned and conditioned before starting of every experiment.

In Fig. 9 the wear rate of the friction material as a function of applied load is shown. It is found that the wear of lining material is increases with applied load and this increase is more for higher loads. Due to the high load the temperature increases at the contact surfaces, there by destroying the transfer film at a faster rate. For all sliding speeds the wear rate of the friction material increases with the applied load. The higher wear rate is because of the presence of the SiC particles in the MMC disc. The SiC particles presence in the counter face destroys the transfer film and the lining material. Initially the SiC particles are sharper and harder, because of the greater degree of sharpness of the SiC particles, the higher amount of stress acts on the pin material, therefore the wear rate of pin material is high. The wear of pin is mainly due to cutting actions of hard SiC particles. With further increase in sliding speed and load the temperature rise increases to a critical value at which the SiC particles act as lubricating agent, thus reduce the wear rate of friction material, also reduces the frictional heating. Fig. 9. Variation of wearin friction material sliding against MMC disc

The wear rate with load for friction material sliding against C.I disc is shown in Fig. 8. At all sliding speeds, the wear rate of friction material is increases with increasing the applied load. Fig. 8. Variation of wear in friction material sliding against cast iron disc

The larger amount of SiC particle attributes the lower wear rate of MMC disc. The reinforcement particles take over the contact stresses and thereby preventing plastic deformations and hence reduce the amount of wear. Fig. 7. Vanation of Wear rate of MMC disc with applied load

Fig. 10. Variation of friction coefficient with applied load sliding against C.I. disc

This is because part of the SiC particles are entered in between the pin and the MMC disc possibly leading to three-body abrasion (tribolayer), resulting in surface roughness between contact surfaces and increase coefficient of friction. These results are in agreement with those reported by Adem Onat [5] and R. N. Rao et al [3]. Fig. 11. Variation of friction coefficient with applied load sliding against MMC disc

During sliding, frictional force acts between the counter surfaces, which cause frictional heating of the counter surfaces. Lower the frictional heating parameter [10], less heat only dissipated to the surroundings. ‘The average value of frictional heating parameter for Al7075- 25wt% SiC composite disc was found to be 9.98x10° C/J, where as it was 14.55x10° ‘CfJ for cast iron disc, therefore the MMC disc mins cooler than C.1I disc. Fig. 12. Increase in contact temperature as a function of load of C.I. disc sliding against friction material

Fig. 13. Increase in contact temperature as a function of load of MMC disc sliding against friction material

Figs. 15. Optical micrographs of friction material surfaces (a) Friction matenal surface before wear (b) Frictional material surface after wear against C.I disc (c) Friction material surface after wear against MMC disc The optical micrographs of friction material before and after wear against cast iron are shown in Figs. 15(a) and 15(b) respectively. Fig. 15(b) shows the wear traces formed on the friction material while sliding against cast iron. It shows the formation of wear grooves and transfer film on the surface of the friction material. The wom surfaces of the friction material while sliding against MMC disc are shown in Fig. 15(c). From Figs. 15(b) and (c) it can be observed that the friction material sliding against cast iron is subjected to less wear damage and with the grooves of smaller width and depth, where as it was observed more for friction material while sliding against composite disc. This is due to the action of hard SiC particles present in the composite disc.

“igs. 14. Optical micrographs of (a) cast iron surface before weat (b) cast iron surface after wear (c) Microstructure of MMC (d) MMC surface after wear The optical micrographs of the contact surfaces of the cast iron, the composite and friction materia are analyzed before and after the wear test at a speed of 7 ms and load of 70 N. Fig. 14(a) shows the contact surface of cast iron before wear test. Fig. 14(b) shows the optical micrograph the cast iron surface after wear test. It shows the wear traces formed on the sliding surfaces.

Figs. 1. Network inputs and outputs to predict (a) cutting forces as a function of time (b) maximum and average cutting forces (c) resultant cutting force To obtain an optimum neural network stnicture to predict the cutting forces in face milling, the effect of varying the type of ANN, the training algorithm, the number of hidden layers, the type of transfer function and the number of neurons per hidden layer must be investigated.

Fig. 2. Network inputs and outputs to predict the cutting coefficients

EFFECT OF TYPE OF TRAINING FUNCTION ON PREDICTION FOR CUTTING CONDITION C4. T1.3. Effect of Varying the Number of Hidden Layers

Fig. 3. Predicted and measured resultant force for one insert cutting, C10 (RPM 1000, Depth of cut of 40 and Chipload of 8) The effect of the type of training algorithm used was investigated next Levenberg-Marquardt (trainlm), Resilient Back Propagation (trainrp) and Scaled Conjugate Back Propagation (trainog) algorithms were To study the effect of varying the number of hidden layers on the accuracy of the ANN prediction, FFBP networks with one and two hidden layes wee investigated. Since the predictions obtained from varying the training function (Table I) showed that trainlm gave the best results, it was used here for all the cases considered. The tansig transfer function was selected as the use of this function allows a nonlinear relationship between the input and the output, moreover, this function is self limiting and hence the output cannot grow infinitely large or small [22]. Table II shows a comparison of the results obtained using one and two hidden layers for cutting condition C8. For the case shown in Table II, the predictions show that a FFBP with two hidden layers result in the lowest error. Additional runs are needed to generalize these findings for other cutting conditions and other types of ANN.

EFFECT OF THE NUMBER OF HIDDEN NEURONS INA FFBP Ty PE ANN WITH ONE HIDDEN LAYER FOR CUTTING CONDITION C10

Fig. 4. Predicted and measured resultant force for one insert cutting, C4 (RPM 1000, Depth of cut of 40, and Chipload of 4)

Fig. 5. Typical comparison between the average cutting force component Fx and the predictions obtained using several neural networks architectures for different cutting conditions

Fig. 6. Typical comparison between the maximum cutting force component Fx and the predictions obtained using several neural networks architectures for different cutting conditions

Fig. 8. Comparison between the calculated values of the radial pressure coefficient and the predictions obtained using various ANN architectures for a variety of cutting conditions Fig. 7. Comparison between the calculated values of the tangential pressure coefficient and the predictions obtained using various ANN architectures for a variety of cutting conditions

RESULTANT FORCES PREDICTION FOR SEVERAL ARCHITECTURES CONSIDERED USING TANSIG/LOGSIG TRANSFER FUNCTION AND TRAINLM TRAINING FUNCTION

Fig. 9. Comparison between the values calculated for the tangential pressure coefficient using the average forces and the predictions obtainec using various ANN architectures for a variety of cutting conditions

Fig. 10. Comparison between the values calculated for the radial pressure coefficient using the average forces and the predictions obtained using various ANN architectures for a variety of cutting conditions

For the cutting conditions considered, the best cutting forces predictions were obtained using FFBP networks with the Levenberg-Marquardt training function algorithm and the tansig transfer function. When predicting the resultant and the maximum forces, the radial basis network produced the lowest error compared to the experimental values. These results are preliminary and additional runs are needed to generalize these findings for other cutting conditions and other ANN architectures. A factorial experiment or a Taguchi experiment can be used for generalization. The approach ensures estimation of the cutting forces in real time which is needed for simulation of different aspects of the machining process such as fixture configuration selection and optimization of cutting In this paper, artificial neural networks were used to estimate the cutting forces and the cutting coefficients usually obtained from mechanistic force mode calibration. It can be claimed that the comparison of the predictions obtained from the neural models with the experimental results confirms the potential of the model to predict both the cutting forces and cutting coefficients.

The main objective for the first stage control is to control the secondary superheater outlet steam temperature at 452 Deg C at rated load. The second stage control type is feedforward cascade PID control and the main objective is to control the boiler outlet steam temperature at 541 Deg C at rated load. The power plant apply 2 stages main steam temperature control circuit as shown in Fig. 2 [1]. The first stage control is feedforward PID control.

Fig. 1. The coal fired power plant steam circuit with important steam temperature setpoint [1]

Fig. 2. Current main steam temperature control circuit [1] Around the same time, Hebb's introduced Hebb's leaming rule or Hebb's synapse [7] which discussed about hypotheses of neural substrate of leaming and memory. However Rosenblatt's work of Perceptron really increase the understanding about artifidal newel network. Rosenblatt's demonstrated his algorithm capability of pattem recognition that could leam and its ability to classify linearly separable pattem classes [8]- [11]. Neural network also used for solving non-linear problem by integrating with other type of non-linear solving method [12]. There are also modeling of the complete power plant utilizing neural network [13].

TABLE II To get better performance from the algorithm, the training and validation data must be adequate and accurate. As such, 26374 sampling data taken from actual plant are taken and fed to the algorithm as the training and validation data for the neural network algorithm. The neural network have four inputs and single output with 2 hidden layers.

The input and output for the propose modeling are as Fig. 3. Result forneural network modeling for primary spray flow control valve using Levenberg- Marquardt leaming algorithm The result showed that the primary spray flow control valve modeling using neural network with Levenberg- Marquardt leaming algorithm able to produce accurate behavior of the control valve when comparing with actual plant. VI. Neural Network with Levenberg- Marquardt Learning Algorithm Structure

This Microcontroller uses a JTAG interface for on- chip-debug and up to 8 MIPS throughput at a8MHzoperating frequency with a 3V supply; it Pulse Width Modulation (PWM) outputs for controlling the robot lights and a serial RS232 port for communication with other devices.

Then, in the second stage a rule base is constructed with the predefined fuzzy mules, which are based on the knowledge about the system characteristics coming from the experts. Since the outputs of fuzzy controllers cannot be used directly, they are converted, in the last stage, to crisp values using a suitable defuzzification method (Fig. 3, [6]). Forward and backward controllers use the whole range of the device universe. Therefore, the maximal values of the emor and its changing range should be equal to the limit of the universe. The universe is in the range of -90° to +90°.

Figs. 4. Membership Functions of the Fuzzy Logic Controller: (a) Membership functions when the robot moves to take a piece froma flat surface, (b) Membership functions when the robot rises up on a 90°ramp, (c) Membership Functions of the speed (PWM) Generally, a 8&bit microcontroller uses triangular shapes and singletons represented in a point-slope format. Singletons require one byte for descriptions while triangular shapes need three bytes: two points located on the variable axis and one slope or grade value. Figs. 4 show the different membership functions used in this work to fuzzifyy the inputs [11].

A. Forward kinematics soludon The implementation was done through MATLAB. The following coding shows the forward solution for the robot:

Fig. 5. Fuzzy PI output surface The Fuzzy PI surface has been plotted using MATLAB in Fig. 5. The PWM output is based on the PI controller equation as Pwm =P +1 [16].

Jy han i. UW. Forx, =0, Xp =0, X3 =0, x4 =1.2, x, =0, we obtain: Copyright © 2014 Praise Worthy Prize S.r.l. - All rights reserved

Fig. 6. Plotting the output values from the microcontroller - Note that the microcontroller integrates many useful capabilities like PWM outputs for controlling the derrick or clamp speed. It is particularly well suited for this type of applications because of its small size and weight and relatively low cost.

Rankine cycle is used to convert waste heat into energy in this waste heat recovery system. The waste from ICE can be obtained in the exhaust manifold and also from the engine radiator system. Fig. 1. Schematic diagram of operation of closed Rankine cycle

Fig. 2. Diagramof ( p—h) forclosed Rankine cycle

SPECIFICATION FOR A PROTON GEN-2 INTERNAL COMBUSTION ENGINE

Fig. 4. Schematic of the experimental setup

Fig. 7. Potential power measured 20 minutes after engine start with rediator fan operating

A. M. Mohd Shafie is a postgraduate Student at Green Technology Vehicles Research Group, Faculty of Mechanical Engineering, Universiti Teknikal Malaysia Melaka His research interest includes heat transfer and intemal combustion engine.

Fig. 1. Touch sensing representation [15] va Sa 6 ESS Skee RRR an RS RE Ip en GR IRE tet The robot stores the position data and then automatically makes adjustments to the weld path prior to the arc start. Touch sensing can make adjustments in one, two, or three dimensions [15]. The robotic process usually starts with the robot in a search start position; a safe voltage is added to either the wire or the gas nozzle (or a probe attached somewhere near the welding torch) and the robot starts moving towards the expected joint position. International Review of Mechanical Engineering, Vol. 8, N. 1

Figs. 2. (a) GITAW robot welding system (b) seam gap viewed by CCD [20] ‘The image is then processed to measure.

Fig. 3. Three directional image acquired inGTAW duminum wadding [14] Tracking and control of the wire extension, thus controlling the wire-feed rate, has been successfully applied to meet the control requirements of wire extension stability [25]. Coupling the two prior sensing cameras permits control of the wire extension and weld pool width at the same time [26]. No application for real-time seam tracking using passive vision has been reported for aluminum GMAW.

Fig. 4. Aluminum boat hull and intemal stiffener structure A URE aS nae 6S TALES Se See aE ~The actual welds are short, because the brackets are welded to both the longitudinal and transversal stiffeners. Modularization process, such as the modular function deployment (MFD), was used directly from the manufacture base. The two MFD factors chosen for consideration were the “common unit” and the “manufacturing process”. In the “common unit’ step, one or several modules were created such that they could be utilized in several boat models, as such or with minor modifications. In the “manufacturing process” step, the module design was considered from the point of view of robotic welding. An old stiffener structure of an aluminum boat, as shown in Fig. 4, was examined from the point of view of robotic welding. Accessibility was somewhat of a challenge for robotic welding due to limited space inside the hull of the aluminum boat. In addition, the large number of triangular brackets made positioning for tack welding rather difficult.

- When welding with typical parameters, a welding speed of 0.9-1.3 m/min was used. In the weldability analysis, it is vital to determine the limits of allowable fit-up errors in the pieces to be joined so that a weld of sufficient quality can be achieved. Based on air gap tests, the largest acceptable tolerance fault on a 3 mm AW 574 sheet is about 1 mm (Figs. 5).

Figs. 6. Optical sensing in welding an aluminum boat, (a) Exterior overlapping joint in the boat bottom sheet, (b) Problems in access to the weld joint inside the boat When welding the inside stiffener structure, problems were encountered because of the lack of space. The structure was so cramped that a welding gun equipped with a large sensor could not fit into the comers between the workpieces (Fig. 6(b)).

The production of biogas depends on different parameters, such as the waste composition, the water content, temperature and density of the medium to be taken into account in the kinetic models hiogas production. The production of biogas changes depending of degradable components waste and quantity. The production of biogas is defined by exponential law proposed by Halvadakis. This law is malfunctioning for the first years degradation of a young waste, but applies verv well to the anaerobic phase (Finidikakis and al) [9]: i ame? I aa Therefore the Eq. (2) and Darcy's law leads to the following equations: where n is the porosity of the medium, and ) , pg the density of liquid and gas phases. V, the rate of filtration of P the phase defined by Darcy's law and ag the term the production of biogas. This latter term is defined from the biological model of degradation and biogas production.

Fig. 1. Saturation profile as a function of time and the axial variable One finds that the volume increases with time, this result can be interpreted in that the component is deposited.

Fig. 4. Volume profile as a function of time and the axial variable VI. Conclusion

Fig. 3(a). Axial Flow Velocity results from CFD Analysis for Combustors with different Swirl Angles For boundary conditions discussed in article II axial velocity profiles for chambers with different swirler

A past researcher shows that, 50° swirl gives the best result in terms of swirl zone and size of recirculation Fig. 4(a). Axial Flow Velocity across section z/R = 0.036 results from CFD Analysis for Combustors with different Swirl Angles

Fig. 3(b). Axial Flow Velocity comparing CFD and experimental results

Fig. 4(b). Axial Flow Velocity across section z/R = 0.072 results from CFD Analysis for Combustors with different Swirl Angles

Fig. 4(c). Axial Flow Velocity across section z/R = 0.143 results from CFD Analysis for Combustors with different Swirl Angles Fig. 4(d). Axial Flow Velocity acress section z/R = 0.286 results from CFD Analysis for Combustors with different Swirl Angles

This can be seen clearly in the Figs. 8(a) to 8(d). The axial flow velocities along the central chamber axis were then plotted until 200 mm (L/D=0.71) from the injection point. In Figure 9, shows that near the injection point, the reverse flow velocities are directly related to the injection velocities. The higher the injection velocities, the faster are the reverse flow velocities. This is accompanied by the reduction in size of the core reverse flow volume. This is understandable since the swirl angle is constant (so is the swirl number), such that the higher injection velocities would produce higher reverse core flow, thus enhancing mixing of fuel in the chamber. Fig. 5(a). Axial Flow Velocity results from CFD Analysis for Combustors with 30° Swirl Angles at different axial distance from the inlet

Fig. 5(b). Axial Flow Velocity results from CFD Analysis for Combustors with 40° Swirl Angles at different axial distance from the inlet

Fig. 5(c). Axial Flow Velocity results from CFD Analysis for Combustors with 50° Swirl Angles at different axial distance from the inlet

Fig. 5(d). Axial Flow Velocity results from CFD Analysis for Combustors with 60° Swirl Angles at different axial distance from the inle'

Fig. 7. Transient flow at 25 s with the injection of various velocities Fig. 6. Axial Flow Velocity results from CFD Analysis for Combustors with 50° Swirl Angles, experiment with 50 m/s air inlet velocity

Fig. 8(a). Axial flow across chamber diameter for injection velocity 30 m/s at different cross section distances from the injection point

Fig. 8(b). Axial flow across chamber diameter for injection velocity 40 m/s at different cross section distances from the injection point Fig. 8(c). Axial flow across chamber diameter for injection velocity 50 m/s at different cross section distances from the injection point

Fig. 8(d). Axial flow across chamber diameter for injection velocity 60 m/s at different cross section distances from the injection point

Fig. 9. Axial flow along the velocities for different

Fig. 1. US electrical energy generation [11] Pollutant emissions are reduced because flame temperatures are typically low, reducing thermal NO, formation. Table I compares the pollutants from natural gas, oil and coal. The use of natural gas will reduce the impact of fossil fuel combustion on climate change. In order to further reduce NO, and other harmful pollutants, lean mixtures will reduce the combustion temperature and decrease the formation of NO,.

Figs. 2. Global greenhouse gas emissions a) by type of gases b) by type of sources [14] Fig. 3 plots the formation of NO,. In order to achieve low NO, emissions, the flame temperature of the combustion must be below 1425°C (1698 K). Above that temperature, the NO,, formation will be very high.

Fig. 3. The rate of NO, formation for flame temperature [13]

Fig. 4. Schematic regime diagram for methane-air JHC flames [39]-[40] The oxygen dilution plays the most important role in achieving MILD combustion as shown in a step by step illustration of oxygen dilution in Figs. 5. Recent applications of MILD combustion have been in research and development of gas turbines [24]-[25], [41] and gasification systems [42]-[43]. This combustion mode can be very interesting in gas turbine applications due to low maximum temperatures (very close to the ones at the inlet of a gas turbine), noiseless characteristics, good flame stability and effectiveness in reducing pollution emissions.

Fig. 6. Carbon dioxide closed cycle for biogas [48] Table III shows a comparison of energy balance for natural gas with 97% methane. The summary was made for the fumace which operates in the flameless mode and conventional mode with natural gas. The supply of thermal energy was constant at about 21 kW for both conditions.

Figs. 5. Step by step of MILD combustion (a) combustion started, (b) to (d) progressively more dilute oxygen (e) fully MILD combustion [45].

The efficiency for the combustion with conventional mode is only 41.4% whereas for MILD mode it is 70%. The comparison for the efficiency of flameless mode and conventional mode for natural gas is 28.6%.

Fig. 8. Efficiency of the heating system with EGR [51]

The difference for Fig. 8 is the fumace running with the regenerator (EGR) and from 654 BTU of heat in the flue gas; only 133 BTU is lost through flue gas to the atmosphere. Some of the 521 BTU of heat is retumed back to the system via the regenerator. The efficiency is 37.4% for the system without EGR and 72.4% for the system with EGR and the system with EGR is 35% higher. Fig. 7. Efficiency of the heating system without EGR [51]

Fig. 9. Industrial fumaces with heat exchanger system [53] The recirculation volume flow into the combustion chamber depends on the level of air pre-heating and oxygen dilution needed. EGR will reduce NO,, emissions of the oxygenated fuels by more than 55% since it reduces both the pressure [52] and the maxinum combustion temperature. Figs. 9 and 10 show the industrial fumace with the heat exchanger system and intemal gas recirculation to utilize the flue gas.

Fig. 10. Industrial closed furnaces with intemal EGR system [54]

The design of the bumer can be used for various purposes. For the USQ combustion group, the experimental setup is not only for MILD combustion but is also applicable for future studies on combustion and ignition including testing the characteristics of altemative fuels under combustion, for example naturel gas, biogas and coal seam gas. Biogas and LCV fuels are difficult to bum in a conventional combustor, but are readily bumed in MILD mode [4], [67], so the potential exists to lead the world in both open fumace MILD systems and the usage of altemative fuels [32], [38]. This also has great potential for consulting work with local industries to improve their green characteristics, and therefore, could lead to substantial future research and development opportunities. Figs. 12.Temperature distnibution for MILD combustion (a) 3D view (b) 2D view

Figs. 11. MILD combustion fumace, (a) University of Southem Queensland, Australia [32] (b) University of Adelaide, Australia [44], [67], [70] (c) Politecnico di Milano, Italy [68] The fumaces for MILD combustion are greatly invested at a laboratory scale and at industrial scale; gradual adaptations are taking place for this new technology. Worldwide there are many research labs and universities are conducting further research, an example of this is shown below in Figs. 11. Fig. 11(a) is MILD combustion in an open fumace at University of Southem Queensland (USQ), Australia [31].

Figs. 2. 2DOF Representation of: (a) Passive Suspension; (b) EMS Suspension The suspension unit consists of a spring and an EMS actuator passive suspension system. The EM actuator was treated as a “black box” that has the ability to produce any amount of force the control system told it to produce. The control strategy will calculate the required force to stabilize the vehicle and feed it to the EM actuator which will supply the required force into vehicle system. For performance comparison, a_ passive suspension system, consisting a passive spring and damper will be used. Figs. 2 show 2DOF models representing the passive suspension setup (a) and EMS suspension setup (b) based on [25]. Here, Frag is the controllable force from EMS actuator.

with the chosen state variables and performance index, further derivations will yield the state space model for EMS as shown in Eq. (7). Here, system input, U is Frag (force to be generated by the EMS actuator), and Z, is the disturbance signal from road input: Copyright © 2014 Praise Worthy Prize S.r.l. - All rights reserved International Review of Mechanical Engineering, Vol. 8, N. 1 Special section on International Conference on Mechanical Engineering Research 2013 (ICMER 2013)

A performance index was chosen next. It consists of the objective functions to be minimized by the LQR contoller. In this study, an established index from [16] is used as shown in Eq. (6) where qi, dp, g3, and qy are the weighted coefficients of each performance velocity (%,), wheel-road displacement (x,-Z,) and

The optimization is solved in MATLAB using Ricatti approach as laid out by [16]. The solution will yield the value of K and thus, the optimum input signal Fymg can be obtained by the optimal control law, Frag=-KX. The Fig Value is the unique solution to optimize the performance index in Eq. (8). A MATLAB/SIMULINK model was built and as shown in Fig. 3. Fig. 3. LOR contoller for EMS system There are several conditions for a system to be successfully controlled by LOR strategy. Therefore, a system check should be carried out to ensure the proposed EMS system fulfilled the conditions as listed below.

Also, performance index from Eq. (6) can be transformed into canonical form in Eq. (8):

A new Set of parameters is introduced as shown in Eq. 0): These effects can be summarised as tabled in Table II. It can be concluded that gg variation will not affect the controller performance and q, should not be zero for stabilized wheel response.

This study wes camied'= out within MATLAB/SIMULINK simulation tools. 2 types of road inputs were used, step and random road input. Step input will be modeling a curb of 0.03m height whereby the random input models a realistic approach of a road surface. In this study, the road surface is generated based on employing the roughest road surface profile (Class E) from ISO standard ([26]). Both road models are shown in Fig. 4. Fig. 4. Step and static random input from ISO 8608 - Class E A set of values were assigned to the weighting parameters in the matrix Q. In this study, the values are: q =1, © =90, gq =10 and q,=Owhich were derived based on guidelines from [16], [22]. The initial values and its subsequent K values calculated are shown in Eq. (9). These values give deteriorating body responses and oscillatory wheel responses. To overcome this, a parametric analysis was camied out to give better understanding on the effect of these values:

Fig. 5. Effect of varying qu compared to passive data for wheel displacement

Fig. 1. Experimental setup of indentation test The failure mechanisms in the FMLs were investigated visually by examined the front and rear surfaces of indented FMLs. Compressive load-indentation depth curves were plotted to analyze their relationship. Also, maximum compressive load for each specimens were identified. The amount of energy absorbed is taken as the area under the load-displacement curve.

Figs. 2. Photographs of the indented (a) pure kenaf fibre plate, (b) 2/1-0.3 FMLs, (c) 2/1-0.6 FMLs, (d) 3/2-0.6 FMLs

Fig. 3. Compressive load versus indentation depth curves for the indentation test

Fig. 4. Maximum compressive load for each specimen

Fig. 5. Energy absorbed of each specimen in indentation test It is interesting to note that before failure of the specimens, the maximum energy absorption of 2/1-0.6 FMLs was 30.82 J (at indentation depth of 9.3 mm) which made it become the highest among the specimens. This is because of debonding failure that occurred in 3/2-0.6 FMLs at the indentation depth of 4.3 mm and it only can achieved the maximum energy absorption of 22.86 J but still higher than 2/1-0.3 FMLs which only 20.14 J at 8.5 mm indentation depth.

Fig. 2. Mechanism of photocatalysis [23] Activation of TiO, is achieved through the absorption of a photon (Av) with ultraband energy from UV irradiation source. This results in the promotion of an electron (e-) from the valence band to the conduction band, with the generation of highly reactive positive holes (h+) in the valence band. This stage is referred as the semiconductor's ‘photo-excitation' state as referred to Fig. 2. The energy difference between the valence band and the conduction band is known as the 'Band Gap. Wavelength of the light necessary for photo-excitation is:

The similar result was found in the IR spectra of water polyaggregates in a nitrogen cryomatrix [21]. Fig. 3. Overlay spectras of FTIR gas cell of BPF;,o of unexposed 60 minutes and exposed to UV light at 60 minutes

It has been reported that the nitrogen doped TiO, could improve visible light adsorption efficiency with characterization of catalyst [27]. Fig. 4. Doping mechanism of nitrogen on titanium dioxide [27]

Figs. 7. Thermal degradation of BPF; o after exposed to UV light at 60 minutes [1]

Fig. 6. Overlay spectra of FTIR for BPF; o before and after exposed to UV light at 60 minutes Fig. 5. OH and C=N intensity of BPF;.0 of exposed at 60 minutes and unexposed to UV light at 60 minutes

In order to achieve a non-invasive inspection, it requires access at both of the tube ends. In this study, the extemal hammer impact is introduced into the tube structure to produce transient stress wave. Extemal impact excitation was introduced into the tube structure using 2kKN instrumented impact hammer BK 8206-002 at one of the tube ends.

é set-up Fig. 3. Experimental A typical time history measurement of the applied mipact hammer for all the tubes are shown in Figs. 4-6.

Hence, the coherence plot from the ten repeated impact is used to validate the quality and consistency of the applied impact hammer force load introduced into the tubes. The coherence plot represents the linear relationship between the impact force load and dynamic response signals. A value near 1 indicates the good quality data. The coherence plots verify that the dynamic response measured by the accelerometer due to the impact force load maintains a linear relationship and thus is consistence for all the impacts. The representative coherence plots are provided in Figs. 4(b)-6(b).

Figs. 6. Vibration impact data of Tube 2 Figs. 5. Vibration impact data of Tube 1 The typical resultant stress wave signals captured by the AE sensor for different type of tubes are depicted in Figs. 7-9. It is clear that all the stress wave signals exhibited a periodical burst type, started at low amplitude until it reached the peak amplitude and followed by repetitive burst signal at gradually lower amplitude before it attenuated.

Based on Table I, the dominant concentration of fatty acids in T.chuii were SFA and MUFA which in total of 81.67%. Bondioli et al., (2012) [20], has reported in the research on oil production from two marine microalgae species Nannochloropsis and Tetraselmis suieca, that the microalgae species T.suecica (similar species to T.chuii) cultivated under nitrogen and phosphate starvation was mainly consist of MUFA (46.98%) followed by SFA (32.76%) and PUFA (18.94%). An important characteristic for any biodiesel feedstock is the suitability of the fatty acid profile for biodiesel production. Table I shows T.chuii lipids were mainly consist of saturated fatty acids C16:0, palmitic acid (23.1%) and C20:0 arachidic (26.2%), followed by monounsaturated fatty acids such as C15:1 cds10- pentadecenoic (7.7%), C18:1 octadecencic (13.0%) and C22:1 erucic (4.0%). Presence of polyunsaturated fatty acids was detected with C22:2, lignoceric (12.6%), C18:2 linoleic (2.7%), C18:3, a-linoleic (2.3%). Other fatty acids (as shown in Table I) were identified with very less amount of percentages.

Log In

International Review of Mechanical Engineering (IREME

Sign up for access to the world's latest research

Abstract

Related papers