Applied Mechanics and Materials Vols. 80-81

Abstract: In the ever-increasing development of network storage technology, distributed RAID has been a hot topic for data reliability and availability. To solve the problem of distributed RAID technology, such as: parallel access difficulty, long rebuild time, small write update and strip alignment problem, we design and implement file level distributed RAID system (BW-FILERAID) based on Blue Whale File System (BWFS). Also we do the research on resource allocation strategy, page cache management with parity data, consistence maintenance, online extension of RAID group, error detect and rebuild and small write optimization in BW-FILERAID. According to the testing results, the writing performance of BW-FILERAID is better than Linux soft RAID about 5%~50%, and the reading performance is equivalent to the RAID0. And it also has excellent extensibility.

Abstract: Numerous resource allocating and releasing will cause fragmentation which will seriously affect the I/O performance of file systems. One method to solve this problem is file defragmentation. This paper focuses on fragmentation detection, the key part of the file defragmentation. We estimate file fragmentation situation through file’s layout, and then analyze the detailed effects of the situation on performance measurement. Then we quantize this by the file’s “fragmentation degree” through theoretical and experimental analyses. Finally we use it to estimate the file’s fragmentation and decide whether the file needs to be defragmented.

Abstract: In this paper, an approach of free space management for FPGA area is proposed. We present a data structure for keeping the information of free space and an algorithm for finding available location to fit the tasks. The simulation proves that the algorithm has higher placement quality and lower complexity than the existed algorithm with certain circumstances.

Abstract: In order to identify and classify insects wing, a new identification algorithms of insects wing was presented in this paper. Wing image was divided into several grids in this algorithm, and according to the importance of the wing features based on every grid, targeted to set the weight for each grid and calculate the final similar, the result of comparison of wing features will be more comprehensive and accurate. Experimental results show that the proposed algorithm will be more accurate to compare wing features, and then identify the insect species.

Abstract: Aiming at issues of the modern embedded turbine supervisory instruments such as simplistic function, weak process power, low efficiency and so on, the technique of double CPUs with the cores of Intel PXA270 and Cyclone II EP2C20 microprocessors is employed in the present design. With embedded linux operating system and the developing platform of Qt software implemented, important functions such as data acquisition, data transfer based on TCP/IP protocols, display of data waveform, local storage of data, data analysis, elementary fault diagnosis and so forth can be achieved all together in a single facility. After tested on the Rotor Kit 4 of Bently Nevada and applied on 300MW turbine of the Power System Dynamic Simulation Laboratory in North China Electric Power University, the newly designed embedded turbine supervisory instruments show that the technique of double CPUs is effective, and the portable system can overcome the deficiencies of the modern embedded turbine supervisory instruments as mentioned above.

Abstract: Fault forecasting model based on dynamic fuzzy synthetically Judging and five units model of forecasting coefficient are built up, and method of selecting coefficient is made certain. On the other hand, this paper put forward technical disposing method of fault judging result, and discusses method of fault pattern recognition. Based on these all establishes important foundation of theory and technology, this system has been successfully adopted in detection, evaluation, and maintenance of large-caliber gun based on states.

Abstract: A phase-locked loop (PLL) control system based on field programmable gates array (FPGA) is proposed through analyzing the model of three-phase unipolar-driven BLDCM (brushless direct current motor) to enhance the reliability and accurate steady-state speed for magnetically suspended control moment gyroscope (MSCMG). The numerical operation module, PLL module and current-loop control module are designed based on FPGA using very-high-speed integrated circuit hardware description language (VHDL) to realize the control law of the digital system. The pulse width modulation (PWM) generating module for Buck converter, the commutation signal generating module for the inverter and ADC module are designed for driving the motor and sampling the current signal. The PLL is analyzed and optimized in the paper and all the modules are verified using the software of ModelSim and the experiments. The simulation and experiment results on BLDCM of MSCMG show that the stability of the motor velocity can reach 0.01% and 0.02% respectively by the PLL technology based on FPGA, which is difficult to be obtained by conventional proportion integral different (PID) speed control.

Abstract: The technique to suppress regenerative chatter is to regulate the spindle speed around the mean speed to disturb the regenerative mechanism. According to the technique, it builds a platform to collect, process the machine working information and output control instruction. On the platform, some advanced algorithms are applied to analyze and finish the control task.

Abstract: The paper is committed to overcome the influence of gross error on the small quantity data of forest fire grey modeling. According to the quantity of the modeling data, Grey judgment of gross error and robust estimation theory is used separately for finding the gross error exit whether or not from the modeling data. And robust estimation theory and LIR algorithm can be used to process the gross error. From the examples, A quarter of fitting precision of robust estimation is less than 1%, and 75% is 1～5%; and half of fitting precision of LIR algorithm is less than 1%, and half is 1～5%. That is to say LIR algorithm provides a rapid, simple and practical way to build model of data which contains gross error or which contain missing data.

Abstract: The model of extracting a skeleton of mechanism is established from an existed three-dimensional CAD product model. Using assembly features, a recomposition of assembly information is carried out. At same time, geometric primitives are defined and classified. And according to the information, kinematic pairs are distinguished and matched. Using the given methods, a prototype system extracting and drawing a skeleton of mechanism is developed. Extracting a skeleton of mechanism from digital mockup has been realized primarily, which can deal with rotation pair, sliding pair, gear pair and multi-pair components consisting of these kinematic pairs. Finally, the paper demonstrates the feasibility of extracting a skeleton of mechanism by an application instances.