Pusat Pengajian Kejuruteraan Mekanikal - Monograf

Browse

Now showing 1 - 5 of 946

Classification analysis for cyclists’ posture recognition
(2016-06)
Chua, Ching Cheng
In cycling sports, there are three basic postures: standing, normal and lean forward to be analyzed for the most efficient posture. Previous works in cycling motion detailed into the physics concept for good cycling performances. However, no studies had realized the body joints and angles postulations worth some consideration. Another constraint is on the complicated steps and processes required for analyzing the cycling motion. Therefore, the purposes of this study are to (i) analyzed three different cycling motion postures: Normal, Lean Forward and Standing for classification analysis, (ii) study skeletal 2D postures that determine optimum posture for classification, (iii) relate selective attributes to enhance the cycling posture classification accuracies. The study begins with experimental cycling motion capture data, followed by data conversion into images and numeric data. Data mining analysis consisting of four stages is employed. It is found that the lower body segments (below waists) are important for classifying data into its corresponding classes. Best classification accuracy obtained show up to 99.67 % accurately classified data using the Tree Classifier on 10 fold cross validation mode. A further study showed that three body joint coordinates: kx, ky and llAngle are the critical attributes, while the knee being the important attribute to distinguish the different cycling postures perfectly.
Investigation of thermal reflow process for copper pillar technology.
(2022-07-25)
Rosli, Muhammad Faiz Ridhwan
Modern gadgets are rapidly growing and providing multiple functions and applications for the users such as smart phones, laptops, desktops, etc. Since the invention of the microchip, modern technology has driven us forward to create smaller gadgets. For older technologies, traditional solder bump was used in most flip chips. Solder bumps have the range size of 75 to 150 microns and copper pillar bump have the range of 50 to 100 microns. Hence, copper pillar technology is taking over the traditional solder bump as it gains popularity to solve the new features and sizes of flip chips. However, the high-quality interconnection joint is dependent on the temperature profile of the reflow soldering process. Therefore, this research paper is pursued to predict the reflow temperature distribution, structure effects with variation of diameter of the solder and copper pillar bump; and validation of the copper pillar bump structure during the reflow soldering. The diameter of the solder and copper pillar bump is the core parameter of this research. This research was done by running a simulation for thermal reflow process by using the ANSYS Fluent software and conducting an experiment using an actual reflow oven with sac305 solder paste and copper substrate. The results show that solder with the diameter of 0.15mm has the highest peak temperature and the lowest peak temperature is the solder diameter of 0.35mm. Other than that, the highest reflow stage starting temperature is the solder diameter of 0.25mm and the lowest reflow stage starting temperature is the solder diameter of 0.35mm. Lastly, the highest final temperature for the thermal reflow process is the solder diameter of 0.35mm and the lowest is the solder diameter of 0.25mm. In order to advance the design for microelectronic industries in the future, this study provides a foundation and insights into the effects of copper pillar bump structure during reflow soldering.
Correction modelling of piezoelectric material non-linear behaviors as an actuator.
(2022-07-24)
Shatmi, Muhammad Syaffi
Piezoelectric material has been widely used as both actuator and sensor in many physical instruments and devices applications, due to its capability and characteristic to react in either way, for example, piezoelectric actuators are used in medical imaging by providing nanometre positioning resolution. However, the undesirable behaviours in piezoelectric material are the nonlinear correlations that subsist in the material to produce the utmost performance. Studies found that some of the nonlinear relations that exist within the piezoelectric actuator are the hysteresis effect, whereby the output displacement of the piezoelectric actuator does not perform an identical reading for each series of measurements. More, the creeping effect also has become an immediate finding whenever most hysteresis loops are measured. Other nonlinear relations such as saturation effect and uncertainty vibration are also found within the piezoelectric actuator. Many approaches have been made to model and also conjugate these nonlinearities effects, especially using different types of mathematical approaches such as the Prandtl-Ishlinski model and Preisach model. These methods are quite complicated rather than using polynomial-based functions where this approach is not only able to model but also inverse the effect of nonlinearity directly. This paper proposed a correction modelling of hysteresis, creep, saturation and uncertainty vibration based on polynomial functions and validate the correction effect through simple a simulation.
Energy-stable residual distribution methods for system of shallow water equations.
(2019-01)
Chang, Wei Shyang
A state-of-the-art Energy-Stable Residual Distribution (ESRD) method is expanded for a system of Shallow Water Equations (SWE) as an improvement over the finite volume counterpart (ESFV) for inheriting multi-dimensional feature, minimal sensitivity to grid distortions and the ability to achieve higher order accuracy with smaller stencil. ESRD imposes energy control simultaneously with the computation of the main variables through the mapping of primary conservative variables to energy variables. The energy conservation and energy stable conditions are achieved via the design of isotropic signals and artificial signals respectively. To preserve the cost-effectiveness of the scheme, the work is limited to only full explicit approach. The main contribution of this work is the source term discretisation which is designed to achieve numerical well-balanceness property. The effects of grid skewness variations on the order of accuracy and stability of ESRD were examined based on scalar analyses. Different degrees of freedom were manipulated to achieve positivity (first order scheme) and linear preserving (second order scheme) properties. A non-linear limited scheme is also constructed with the blending of the first and second order schemes. Unlike ESFV, ESRD demonstrates its ability to preserve the order of accuracy even on high randomized triangular grids. The well-balancedness of the proposed scheme was validated numerically and the order of accuracy of the well-balanced version of the schemes are still preserved.
Forecasting driving distraction using data mining analysis.
(2022-07)
Mohd Nazrol, Nurul Atikah
Distracted driving poses significant safety risk factors to road users. The source of distractions varies from visual, manual, auditory also cognitive. Research has been devoted to analyzing driving distraction based on secondary tasks, mobile phones, and type of roadway effect. However, insufficient analysis was carried out to forecast driving distraction levels. Few studies have considered the effect of gender and the relationship between driving performance measures. Therefore, this research aims to identify potential metrics that indicate driving distraction state, determine gender effect on driving distraction levels., and create a classification model that forecasts driving distraction levels based on driving performance metrics. The case study datasets that involve eight sessions of experimental driving on a simulator were retrieved from the Open Science Framework public databases. Preprocessing techniques were applied to eliminate outliers, extreme values, and irrelevant instances. The datasets are trained on 10-fold cross-validation classification using all the algorithms to determine three distracted driving levels: low, moderate, and high. The classification performances are compared with the ZeroR baseline algorithms. Outcomes show that the percentage accuracy obtained from the OneR classification model performed best, giving 99.6674% % accuracy. Findings also show that female is prone to distraction compared to male.

Browse

Recent Submissions