ACS Applied Computer Science

  • Increase font size
  • Default font size
  • Decrease font size

Applied Computer Science Volume 17, Number 3, 2021

PERFORMANCE ENHANCEMENT OF CUDA APPLICATIONS BY OVERLAPPING DATA TRANSFER AND KERNEL EXECUTION

Print

The CPU-GPU combination is a widely used heterogeneous computing system in which the CPU and GPU have different address spaces. Since the GPU cannot directly access the CPU memory, prior to invoking the GPU function the input data must be available on the GPU memory. On completion of GPU function, the results of computation are transferred to CPU memory. The CPU-GPU data transfer happens through PCI-Express bus. The PCI-E bandwidth is much lesser than that of GPU memory. The speed at which the data is transferred is limited by the PCI-E bandwidth. Hence, the PCI-E acts as a performance bottleneck. In this paper two approaches are discussed to minimize the overhead of data transfer, namely, performing the data transfer while the GPU function is being executed and reducing the amount of data to be transferred to GPU.  The effectiveness of these approaches on the execution time of a set of CUDA applications is realized using CUDA streams. The results of our experiments show that the execution time of applications can be minimized with the proposed approaches.

BREAST CANCER DIAGNOSIS USING WRAPPER-BASED FEATURE SELECTION AND ARTIFICIAL NEURAL NETWORK

Print

Breast cancer is commonest type of cancers among women. Early diagnosis plays a significant role in reducing the fatality rate. The main objective of this study is to propose an efficient approach to classify breast cancer tumor into either benign or malignant based on digitized image of a fine needle aspirate (FNA) of a breast mass represented by the Wisconsin Breast Cancer Dataset. Two wrapper-based feature selection methods, namely, sequential forward selection(SFS) and sequential backward selection (SBS) are used to identify the most discriminant features which can contribute to improve the classification performance. The feed forward neural network (FFNN) is used as a classification algorithm. The learning algorithm hyper-parameters are optimized using the grid search process. After selecting the optimal classification model, the data is divided into training set and testing set and the performance was evaluated. The feature space is reduced from nine feature to seven and six features using SFS and SBS respectively. The highest classification accuracy recorded was 99.03% with FFNN using the seven SFS selected features. While accuracy recorded with the six SBS selected features was 98.54%. The obtained results indicate that the proposed approach is effective in terms of feature space reduction leading to better accuracy and efficient classification model.

APPLICATION OF A FUZZY CONTROLLER IN THE PROCESS OF AUTOMATED POLYETHYLENE FILM THICKNESS CONTROL

Print

The present article aims to describe the design of a fuzzy controller used for automated control of the thickness of the extruded polyethylene film effected by the adjustment of the actuator in the cooling ring. In order to determine whether the designed controller operates properly, a model extruder was created and a simulation study was carried out. The Simulink programming environment integrated with Matlab was used for the development of the fuzzy controller and the simulation. The conducted simulation study demonstrated that the implementation of the designed controller would enable the adjustment of thickness on the perimeter of the film tube and quick reaction to possible departure in the assumed film thickness in mass production.

OPTIMAL SLIDING MODE CONTROLLER DESIGN BASED ON WHALE OPTIMIZATION ALGORITHM FOR LOWER LIMB REHABILITATION ROBOT

Print

The Sliding Mode Controllers (SMCs) are considered among the most common stabilizer and controllers used with robotic systems due to their robust nonlinear scheme designed to control nonlinear systems. SMCs are insensitive to external disturbance and system parameters variations. Although the SMC is an adaptive and model-based controller, some of its values need to be determined precisely. In this paper, an Optimal Sliding Mode Controller (OSMC) is suggested based on Whale Optimization Algorithm (WOA) to control a two-link lower limb rehabilitation robot. This controller has two parts, the equivalent part, and the supervisory controller part. The stability assurance of the controlled rehabilitation robot is analyzed based on Lyapunov stability. The WO algorithm is used to determine optimal parameters for the suggested SMC. Simulation results of two tested trajectories (linear step signal and nonlinear sine signal) demonstrate the effectiveness of the suggested OSMC with fast response, very small overshoot, and minimum steady-state error.

BACKWARD MOTION PLANNING AND CONTROL OF MULTIPLE MOBILE ROBOTS MOVING IN TIGHTLY COUPLED FORMATIONS

Print

This work addresses the development of a distributed switching control strategy to drive the group of mobile robots in both backward and forward motion in a tightly coupled geometric pattern, as a solution for the deadlock situation that arises while navigating the unknown environment. A generalized closed-loop tracking controller considering the leader referenced model is used for the robots to remain in the formation while navigating the environment. A tracking controller using the simple geometric approach and the Instantaneous Centre of Radius (ICR), to drive the robot in the backward motion during deadlock situation is developed and presented. State-Based Modelling is used to model the behaviors/motion states of the proposed approach in MATLAB/STATEFLOW environment. Simulation studies are carried out to test the performance and error dynamics of the proposed approach combining the formation, navigation, and backward motion of the robots in all geometric patterns of formation, and the results are discussed.

RGB-D FACE RECOGNITION USING LBP-DCT ALGORITHM

Print

Face recognition is one of the applications in image processing that recognizes or checks an individual's identity. 2D images are used to identify the face, but the problem is that this kind of image is very sensitive to changes in lighting and various angles of view. The images captured by 3D camera and stereo camera can also be used for recognition, but fairly long processing times is needed. RGB-D images that Kinect produces are used as a new alternative approach to 3D images. Such cameras cost less and can be used in any situation and any environment. This paper shows the face recognition algorithms’ performance using RGB-D images. These algorithms calculate the descriptor which uses RGB and Depth map faces based on local binary pattern. Those images are also tested for the fusion of LBP and DCT methods. The fusion of LBP and DCT approach produces a recognition rate of 97.5% during the experiment.

IMPLEMENTATION OF DYNAMIC AND FAST MINING ALGORITHMS ON INCREMENTAL DATASETS TO DISCOVER QUALITATIVE RULES

Print

Association Rule Mining is an important field in knowledge mining that allows the rules of association needed for decision making. Frequent mining of objects presents a difficulty to huge datasets. As the dataset gets bigger and more time and burden to uncover the rules. In this paper, overhead and time-consuming overhead reduction techniques with an IPOC (Incremental Pre-ordered code) tree structure were examined. For the frequent usage of database mining items, those techniques require highly qualified data structures. FIN (Frequent itemset-Nodeset) employs a node-set, a unique and new data structure to extract frequently used Items and an IPOC tree to store frequent data progressively. Different methods have been modified to analyze and assess time and memory use in different data sets. The strategies suggested and executed shows increased performance when producing rules, using time and efficiency.

INNOVATIVE DEVICE FOR TENSILE STRENGTH TESTING OF WELDED JOINTS: 3D MODELLING, FEM SIMULATION AND EXPERIMENTAL VALIDATION OF TEST RIG – A CASE STUDY

Print

This work shows a case study into 3D modelling, numerical simulations, and preliminary research of self-designed test rig dedicated for uniaxial tensile testing using pillar press. Innovative device was CAD modelled, FEM optimized, build-up according to the technological documentations. Then, the device utilization for tensile testing was validated via preliminary research. 3D model of the device was designed and FEM-analyzed using Solid Edge 2020 software. The set of FEM simulations for device components made of structural steel and stainless steel and at a workload equal 20 kN were conducted. This made it possible to optimize dimensions and selection of material used for individual parts of the device structure. Elaborated technical documentation allows for a build-up of a device prototype which was fixed into the pillar press. After that, the comparative preliminary experiments regarding tensile strength tests of X5CrNi18-10 (AISI 304) specimens were carried out. Tests were done using the commercial tensile strength machine and obtained results were compared with those received from an invented device. The ultimate tensile strength of X5CrNi18-10 steel, estimated using the commercial device (634 MPa) and results obtained from the patented device (620 MPa), were in the range of the standardized values. Findings confirm the utilization of the invented device for tensile strength testing.