Genetic optimization for engine combustion system calibration: A case study of optimization performance sensitivity to algorithm search parameters

The use of genetic optimization algorithms (GOA) has been shown to significantly reduce the resource intensity of engine calibration, motivating investigation into the development of these methods. The objective of this work was to quantify the sensitivity of GOA performance to the algorithm search parameter values, in a case study of engine calibration. A GOA was used to calibrate four combustion system control parameters for a direct-injection gasoline engine at a single operating condition, with an optimization goal to minimize brake specific fuel consumption (BSFC) for a specified engine-out NOx concentration limit. The calibration process was repeated for two NOx limit values and a wide range of values for five GOA search parameters, including the number of genes, mutation rate, and convergence criteria. Results indicated GOA performance is very sensitive to algorithm search parameter values, with converged calibrations yielding BSFC values from 1 to 14% higher than the global minimum value, and the number of iterations required to converge ranging from 10 to 3,000. Broadly, GOA performance sensitivity was found to increase as the NOx limit was decreased from 4,500 to 1,000 ppm. GOA performance was the most sensitive to the number of genes and the gene mutation rate, whereas sensitivity to convergence criteria values was minimal. Identification of one set of algorithm search parameter values which universally maximized GOA performance was not possible as ideal values depended strongly on engine behavior, NOx limit, and the maximum level of error acceptable to the user.

An Optimized Feed Hexagonal Antenna with Defective Ground Plane for UWB Body Area Network Application

Body area network has facilitated monitoring, authentication and security through sensors or microstrip antenna with specified frequency. The purpose of this research work is to propose a simplified way to search for an optimal length of the inset for edge feeding using the Evolutionary Algorithm search by minimizing the reflection coefficient using ANSYS HFSS. The optimal inset length resulted in an antenna with better radiation efficiency and wider bandwidth. The antenna structure is 70x70x1.6 mm3, with a modified ground. The purpose of this antenna is communication in Ultra-wideband and works in 5.4, 8.1, and 9.8 GHz bands respectively. The resonant bandwidth measured are 1.02, 0.28, 0.12 GHz, respectively. Simultaneously the achievable gains are 3.18, 7.81, and 19.95 dB, respectively in free space. As the antenna is of wearable type, the front-to-back ratio evaluated for each band is 2.31, 7.01, and 13.91 respectively. The results of the fabricated antenna agree with the simulated results. The specific absorption rates at resonant frequencies were observed to be 0.3, 0.56 and 0.24 W/kg respectively when antenna is placed on a human tissue model. The antenna is useful for on-body communication at ISM band, and high data rate off-body communication in body area networks.

VIDEO ANALYSIS AND 3D DETECTION SIMULATION OF JUMP SHOT PRECISION FOR BASKETBALL PLAYERS

ABSTRACT With the rapid development and application of computer technology, the application of computer science knowledge in basketball is also more and more extensive. Based on genetic algorithm and the background subtraction method, video analysis and 3D detection simulation model of shot jump action precision were constructed in this study. According to the genetic algorithm search method, jump shot precision was analyzed, and the problems encountered in the actual shooting process of basketball players were studied and solved. The results show that this study is necessary and feasible.

The Classification of Public Complaints through Smart City Application

This research aimed to classify the data of public complaints of people in Tangerang City by applying a pattern of the complaint data from the LAKSA application that has been categorized. In finding the pattern, it used one of the data mining methods, namely classification. The classification algorithm search process was performed by comparing the accuracy of several selected algorithms. The algorithms were k-nearest neighbor, random forest, support vector machine, and AdaBoost. These algorithms were tested to achieve maximum potential. Thus, the results showed support vector machine with linear kernel is a classification algorithm with the highest accuracy that reached 89.2%