Advanced Materials Research Vols. 139-141

Abstract: In this paper, the problem of global asymptotic stability in the mean square for stochastic fuzzy cellular neural networks (SFCNN) with time-varying delays is investigated. By constructing a newly proposed Lyapunov-Krasovskii function (LKF) and using Ito’s stochastic stability theory, a novel delay-dependent stability criterion is derived. The obtained stability result is helpful to design the stability of fuzzy cellular neural networks (FCNN) with time-varying delays when stochastic noise is taken into consideration. Since it is presented in terms of a linear matrix inequality (LMI), the sufficient condition is easy to be checked efficiently by utilizing some standard numerical packages such as the LMI Control Toolbox in Matlab. Finally, an illustrate example is given to verify the feasibility and usefulness of the proposed result.

Abstract: The partitioned iterated function systems (PIFS) were introduced into the compression of vibration signal. The actual vibration data of ZLH600-2 pump were adopted to verify the performance of PIFS. Also the compression ratios and the computation time were good, but the spectrum and amplitude, as important performances, were deformed after compression. If the concerned frequency of users was significantly lower than the sampling frequency and the required compression ratios was not more than 2, the compression using PIFS in vibration signal of rotating machinery was comfortable. Otherwise, the information loss in the compression could not be ignored. The decoded signals with the different compression parameters were listed in the paper and it was a meaningful exploration of the IFS on diagnosis and laid the foundation for further research.

Abstract: As a result of the uneven distribution of the hardness of coal roadway section, the intelligent tunneling of roadheader is very important for cutting efficiency of roadheader in various working conditions. Profile cutting is the key technology of realizing intelligent tunneling of roadheader, and the trajectory planning method of cutting head is the technical bottleneck of solving profile cutting. In consideration of the randomness of cutting process, this paper designs specially a time and speed predictor by adopting the spatial interpolation method of joint, considers at the same time cutting-motor as a constraint condition, and uses radial basic network method to approximate the trajectory of cutting head. A novel and effective method is provided for profile cutting technology of roadheader.

Abstract: For the nine characteristic factors of tobacco leaf grading standards have different degree of influence on final grading results and lack of objective evaluation method, in this paper, we applied the gray relational analysis method to determine the weight of tobacco leaf factors in every grade, which calculate the gray relational analysis of nine characteristic factors: such as hue, lightness value, chroma, length, leaf structure, waste, oil, maturity and body. The gray relation was normalized to get the weight of the nine factors in tobacco leaf classification. By contrasted with the subjective evaluation of five experts in tobacco field, the calculation results are basically consistent with the experts’ recommendation. It illustrates that the application of Grey relational method to calculate influence ability of flue-cured tobacco grading factors is feasible. This method eliminates the subjectivity of the weight of each factor and can make the results more realistic.

Abstract: This paper presents One-Distributed-Multi-Agent-Plan. In this system, each agent is planning to carry out their goals and bound by the plan and to determine action steps. Action steps with the routine sequence of steps bound to the time sequence Equivalence of constraints and variables used restraint is not equivalent to not entirely routine steps; Programming is gradually adding and refining constraints on the planning process, algorithm based on the classic UCPOP Planning algorithm; through a special Multi-Agent arrange to check and clear up collision between Agent-Plan, then restriction their action. In this algorithm, The coordination and Conflict-Detection of the Agent-Plan consulted by Multi-Agent I.e.; the bound consistency of Distributed judgment between Agent-Plan; The algorithm is reliable in determining the environment, Since the algorithm is exchanged between agents and actions related to the conflict, and the causal chain bound Therefore it is a small amount of communication and high security advantages.

Abstract: We present an approach to recognizing characters in surface mount technology (SMT) product. An improved SMT product character recognition method is proposed which can obtain a good recognition rate. Some appropriate image processing algorithms, such as Gray processing, Low-pass Filter, Median Filter, and so on, are used to eliminate the noise. Then, Character image is obtained after character segmentation and character normalization. Finally, a three-layer back propagation (BP) neural network module is constructed. In order to improve the convergence rate of the network and avoid oscillation and divergence, the BP algorithm with momentum item is used. As a result, the SMT product character recognition system is developed. Experimental results indicate that the proposed character recognition can obtain satisfactory character-recognition rate and the recognition rate reached over by 98.6% when the hidden layer of BP neural network module has 20 nodes.

Abstract: Prediction of stroke for precision straightening is core part of flexible automatic straightening machine. Straightening process is a nonlinear process, and affected by many factors. In this paper, in order to improve the straightening precision, an expert system is developed based on the previous research, which mainly consists of database for keeping data, expert program for self-study and procedures for application. This paper focuses on designing the expert system and application process. Humanization of the software interface has been designed for ordinary operators and researchers. And at last this system’s feasibility is verified by comparison between these two results, the performance of established expert system is excellent. It will be helpful to the development of automatic straightening machine’s control system.

Abstract: The error measuring, modeling and compensation techniques for the positioning stage driven by NC linear motors are studied. The error source of the positioning stage is analyzed, the positioning errors are measured by the laser interferometer, and the neural network error model is set up by RBF algorithm. In order to evaluate the accuracy of RBF network prediction method, part of the error samples are used to test. A DSP-core linear motor experimental platform is built up, the error compensation experiments are conducted, the real-time requirement is proved to be met. The simulation and experimental results indicate that the RBF neural network error model trained by samples has a good learning ability and generalization ability, the positioning accuracy is improved significantly, and the effect of random errors on the system is reduced also.

Abstract: In this paper a new control method has been studied in which PID control system was integrated into the neural network. It could overcome some disadvantages such as neural network’s slow rate of convergence and PID’s difficulty in application of multivariate nonlinear systems. A controller of the Electro-hydraulic proportional control stroking mechanism for radial piston pump was designed based on the PID neural network control algorithm. The system responses of system variable control signal of system track were achieved by computer simulation. It was found by PIDNN that the control system could reach steady state in a shorter time, compared with PID control system response time by 65% to 80%. The simulation results showed that the controller for the Electro-hydraulic proportional Radial Piston Pump based PID neural network control algorithm would have a good controlling performance.

Abstract: Nowadays the production of mold seriously restricts the manufacture of products as well as the development of new products, it has become an urgent problem to be solved. The paper mainly discussed the fuzzy neural network model and learning algorithm, and utilized expert evaluating system to acquire the training and test samples. Moreover, it established the related mapping model for fuzzy neural network to evaluate the assemblability of mold, so as to improve the productivity of mold. By adopting two different fuzzy neural networks to contrast and evaluate the assemblability evaluation system of the parts of windshield mold, it was concluded that the improved fuzzy neural network model had advantage over the conventional one. Finally, the satisfactory results of assemblability evaluation system of windshield mold had been achieved by coming with examples to carry out error analysis of the assemblability evaluation system.