Kategori: SCI & SCI Expanded

Within the scope of this study, a Field Programmable Gate Array (FPGA) based system which calculates bearing angles by analyzing the signals emitted by vessels in underwater environment is proposed. An array consisting of 3 non-directional hydrophones were designed and used in tests. Minimum Variance Distortionless Response (MVDR) algorithm was used in the bearing calculations according to the reference hydrophone. In marine tests, hydrophone array was integrated into the buoy, and then tested. The angle between the reference hydrophone and the magnetic north is calculated in order to correct the errors caused by the immobility of the buoy and underwater array. In the developed system, all operations were carried out on Artix 7 FPGA. Fixed point number format is used and implementation stages are designed as pipeline architecture. In the marine tests performed, it was monitored in real time that the bearing information calculated by the system was compatible with the route of the vessel used in the tests. The signals received by bearing information and the hydrophones were recorded. The records were run offline, and the calculated values were compared. The results obtained showed that the developed FPGA-based system successfully calculates the bearing angle of the vessels by passive listening.

The main purpose of DNS is to convert domain names into IPs. Due to the inadequate precautions taken for the security of DNS, it is used for malicious communication or data leakage. Within the scope of this study, a real-time deep network-based system is proposed on live networks to prevent the common DNS tunneling threats over DNS. The decision-making capability of the proposed system at the instant of threat on a live system is the particular feature of the study. Networks trained with various deep network topologies by using the data from Alexa top 1 million sites were tested on a live network. The system was integrated to the network during the tests to prevent threats in real-time. The result of the tests reveal that the threats were blocked with success rate of 99.91%. Obtained results confirm that we can block almost all tunnel attacks over DNS protocol. In addition, the average time to block each tunneled package was calculated to be 0.923 ms. This time clearly demonstrates that the network flow will not be affected, and no delay will be experienced in the operation of our system in real-time.

Due to their chaotic nature, underwater communication channels contain many adverse factors affecting the communication link quality and its performance. These adverse effects directly affect the data transfer between the source and the receiver. Absorption loss, which is one of these adverse factors, depends on depth, temperature, salinity, pH, and speed of sound, as well as frequency, and it has direct impact on the bandwidth used by the system and the distance required for reliable communication. In this study, the effects of variation of temperature, salinity, depth, and sound velocity on the channel bandwidth, channel capacity, and transmission power of the channels formed in the underwater environment in Erdek/Turkey were examined. Within the scope of the study, estimations of the bandwidth, capacity and transmission power parameters were conducted by using temperature, salinity, and sound velocity data relative to the depth recorded between July 2018 and December 2018. Cylindrical, spherical, and practical propagation models are used to compute the propagation loss. In contrast to the studies performed in the literature regarding absorption loss calculations, instead of using only the frequency-dependent approach, realistic models were created by including the effect of changes in the underwater environment in the channel estimation calculations using measurement data. Simplified absorption loss parameters for absorption loss calculations are proposed in the study. It was observed that the channel estimated within the scope of the study are compatible with the outputs obtained from the analysis.

Artificial Neural Network (ANN) training using gradient-based Levenberg & Marquardt (LM) algorithm has been implemented on FPGA for the solution of dynamic system identification problems within the scope of the study. In the implementation, IEEE 754 floating-point number format has been used because of the dynamism and sensitivity that it has provided. Mathematical approaches have been preferred to implement the activation function, which is the most critical phase of the study. ANN is tested by using input-output sample sets, which are shown or not shown to the network in the training phase, and success rates are given for every sample set. The obtained results demonstrate that implementation of FPGA-based ANN training is possible by using LM algorithm and as the result of the training, the ANN makes a good generalization.

In this study, field-programmable gate array (FPGA)-based hardware implementation of the wavelet neural network (WNN) training using particle swarm optimization (PSO) and improved particle swarm optimization (iPSO) algorithms are presented. The WNN architecture and wavelet activation function approach that is proper for the hardware implementation are suggested in the study. Using the suggested architecture and training algorithms, test operations are implemented on two different dynamic system recognition problems. From the test results obtained, it is observed that WNN architecture generalizes well and the activation function suggested has approximately the same success rate with the wavelet function defined in the literature. In the FPGA-based implementation, IEEE 754 floating-point number format is used. Experimental tests are done on Xilinx Artix 7 xc7a100t-1csg324 using ISE Webpack 14.7 program.

This paper presents the first hardware implementation of neuro-fuzzy system (NFS) with its metaheuristic learning ability on field programmable gate array (FPGA). Metaheuristic learning of NFS for all of its parameters is accomplished by using the improved particle swarm optimization (iPSO). As a second novelty, a new functional approach, which does not require any memory and multiplier usage, is proposed for the Gaussian membership functions of NFS. NFS and its learning using iPSO are implemented on Xilinx Virtex5 xc5vlx110-3ff1153 and efficiency of the proposed implementation tested on two dynamic system identification problems and licence plate detection problem as a practical application. Results indicate that proposed NFS implementation and membership function approximation is as effective as the other approaches available in the literature but requires less hardware resources.

This paper presents an approach for modeling and prediction of both surface roughness and cutting zone temperature in turning of AISI304 austenitic stainless steel using multi-layer coated (TiCN + TiC + TiCN + TiN) tungsten carbide tools. The proposed approach is based on an adaptive neuro-fuzzy inference system (ANFIS) with particle swarm optimization (PSO) learning. AISI304 stainless steel bars are machined at different cutting speeds and feedrates without cutting fluid while depth of cut is kept constant. ANFIS for prediction of surface roughness and cutting zone temperature has been trained using cutting speed, feedrate, and cutting force data obtained during experiments. ANFIS architecture consisting of 12 fuzzy rules has three inputs and two outputs. Gaussian membership function is used during the training process of the ANFIS. The surface roughness and cutting zone temperature values predicted by the PSO-based ANFIS model are compared with the measured values derived from testing data set. Testing results indicate that the predicted surface roughness and cutting zone temperature are in good agreement with measured roughness and temperature.

This work introduces hardware implementation of artificial neural networks (ANNs) with learning ability on field programmable gate array (FPGA) for dynamic system identification. The learning phase is accomplished by using the improved particle swarm optimization (PSO). The improved PSO is obtained by modifying the velocity update function. Adding an extra term to the velocity update function reduced the possibility of stucking in a local minimum. The results indicates that ANN, trained using improved PSO algorithm, converges faster and produces more accurate results with a little extra hardware utilization cost.

In this paper, two-layered feed forward artificial neural network’s (ANN) training by back propagation and its implementation on FPGA (field programmable gate array) using floating point number format with different bit lengths are remarked based on EX-OR problem. In the study, being suitable with the parallel data-processing specification on ANN’s nature, it is especially ensured to realize ANN training operations parallel over FPGA. On the training, Virtex2vp30 chip of Xilinx FPGA family is used. The network created on FPGA is coded by using VHDL. By comparing the results to available literature, the technique developed here proved to consume less space for the subjected ANN training which has the same structure and bit length, it is shown to have better performance.