Applied Mechanics and Materials Vols. 20-23

Abstract: Association rules provide a concise statement of potentially useful information, and have been widely used in real applications. However, the usefulness of association rules highly depends on the interestingness measure which is used to select interesting rules from millions of candidates. In this study, a probability analysis of association rules is conducted, and a discrete kernel density estimation based interestingness measure is proposed accordingly. The new proposed interestingness measure makes the most of the information contained in the data set and obtains much lower falsely discovery rate than the existing interestingness measures. Experimental results show the effectiveness of the proposed interestingness measure.

Abstract: The traditional OCR obtains unsatisfactory results in the field of image recognition when images are processed in a complex background with low quality. This paper presents a novel application of the model of Bag of Words on Chinese character recognition, and extensively evaluated its effectiveness with 12 different fonts of Chinese character datasets under varying circumstances. Our experimental results demonstrate that this approach can achieve nearly 70% at its highest accuracy rate, which shows its performance far exceeds the traditional OCR’s

Abstract: Estimation of curvature on digital curves plays a critical role for various applications in the field of image processing and computer vision. The refined L-curvature has proven to be an efficient measure of discrete curvature for corner detection and curve matching. In this paper its stability under rotation transformations is evaluated. Experiments with comparisons to other existing curvature measures show that the refined L-curvature performs best.

Abstract: In connection with the situation that it is difficult to measure high-pressure flow in hydraulic system. A new type liquid flowmeter was designed in this paper, shown in Fig.1. The gear flowmeter is the positive displacement flowmeter with good features that the measurement precision is very accurate, and it is not restricted by environmental condition. But the gear flowmeter can't be widely used, just because its flow pulsation is large. After deeply studying the gear flowmeter, the author designed a new planet gear flowmeter. The flowmeter has the characteristics that the displacement is large, and the flow pulsation is small, even less than 1%. It can also be used to measure the flow in high pressure hydraulic system.

Abstract: Gas turbine machine,Multi-level recursive,Prediction,Experimental study Abstract. Take the huge gas turbine set of Beijing Yanshan Petrochemical Refinery as the research object. The most important feature of machine is the vibration characteristic,and vibration displacement is the most important parameter of the machine,based on the historical vibration displacement datas,use multi-level recursive method to predict the state of the machine.With this method,fault prediction is divided into two parts:forecast of time-varying parameters and forecast based on the former forecast,the prediction precision is improved. Through the analysis of gas turbine failure mechanism,use Bently rotor bench,effectively simulate the mechanical failure of the rotor imbalance.

Abstract: Traditional simulation systems were capable of providing statistical evaluation of targeted factory design feasibility. But designers cannot experience the actual setting of factory design and simulation. This paper illustrates the implementation of virtual factory layout system in 3D virtual environment. Two key techniques and functionalities are presented and visualized in our system. Based on the model database, designers can select proportionate physical object from database and interactive layout them in virtual factory. Furthermore, we define the logical model and establish the communication between logical models for simulation.

Abstract: In this paper we consider the polling system with multiple vacations an bulk arrival using M-gated services in continuous time. By the imbedded Markov chain theory, the generating functions of queue length at the station is obtained. Then computational equations are explicitly determined for the mean queue length. Especially we can obtain some corresponding results under some especial cases. The results reveal that our system model can guarantee better QoS and system stability, and it has better efficiency than that of traditional gated service.

Abstract: The individual decision-making is unstable and risky, comparing with group decision-making. The latter can reduce the risks arising from the decision-making through bringing together all aspects of information. This article introduces a method of group decision-making based on fuzzy evaluation information, and improves the feasibility and exactness of current evaluation models. It takes job value as the target and evaluates the relative value of four jobs. The results show that the model is simple and practical except of its limited availability.

Abstract: This paper proposed an algorithm for license plate recognition system(LPRS). The vertical edge was first detected by sobel color edge detector. Then, the invalid edge was removed regarding edge density. Next, the license plate(LP) image was converted into HSV color model, and by edge density template and fuzzy color information judgement, the LP region was located. Then, color-reversing judgement and tilt correction was conducted. Afterward, characters were segmented by means of vertical projection and convolution, by which character width and position can be exactly confirmed, and character recognition was conducted based on radial basis function (RBF) neural network. With a lot of samples verified in night hours and daytime under real conditions, the experimental results show that the proposed method can achieve accuracy and effectiveness in LPRS.

Abstract: To overcome inherent resolution limitation of imaging system and provide a high resolution satellite image economically and practically, the paper researched on image super-resolution reconstruction (SRR) based on BPNN and RBFNN. Learning sample images are obtained according to image observation model, input/output mapping pairs are formed by way of vector mappings, and then two NN models of SRR are established respectively. Simulation and generalization tests are carried on the BPNN and RBFNN respectively. Results indicate that reconstruction performance of BPNN is better than that of RBFNN in aspect of reconstruction time-consuming, convergence speed and effect of SRR in generalization tests, although those parameters are better in simulation tests of RBFNN. So BPNN is more suitable for project applications of image super-resolution reconstruction.