 Wysłany: Nie 3:17, 08 Maj 2011 Temat postu: franklin et marshall BP neural network based non-s |
|
| ,franklin et marshall
BP neural network based non-stationary excitation mechanical system response predicted 50kg, I】 = 13600kgm. , A 12300kg, I2-21560kgm,tory burch outlet, Ma = a sleep = M a 93okg, q-C2-1000N · s / m,Franklin Marshall pas cher, k7 k8 a k9 a one kIo = 53500kN / m; d = 2.2m; 6 a o. 346m; d a 3.3m; a 1.96m; = 2.14m; lost human {k3kkk) and {,,} of these eight systems such as the design parameters, the rms output is 2,0. Corresponding BP network input layer of 8 units, the first hidden layer of 9 units, the second layer of 4 hidden units, output is 2 units. Taking into account the axis of the same sex, the design parameters k = one hundred and eleven; one by one for a c a f c; of 10 sets of samples to train the network. In another five input spreadsheet, BP network approximation of the value of the results compared with the Newmark method list 1 below. Table 1BP network calculation results 5 Conclusion The four non-stationary excitation BP network under the condition of the mechanical system response assessment, dynamic optimization, such as the need to repeatedly solve the same system under dynamic equations, the equation can be used a few times Solving the network training phase 2 Li Qiang et al: The network based on BP stretch of non-stationary response of the mechanical system tender Xu Li estimated 199 training, assessment and efficient response can be mapped, so has a unique advantage. For a particular problem, many judges computing time (excluding the study, asked) many times over Newmark method for solving time is much shorter. However, the general purpose of the poor, for different issues need to re-learn. References THEORETICAL so on. Structure of the mixed non-stationary random response of precise integration platform type. Vibration Engineering, 1995f8 (2) :127-134 Dong Cong,Christian Louboutin, etc. Progress in multi-layer network segment to a number of issues. Mechanics, l995 {25 (2): 186 a l96 Jiao Licheng. Stretch by the network theory. Xi'an: Xidian University Press,Oakley juliet, 1993RumelhartDE, H [ntonGEtWilliamsRJ. LearningrepresentationbybackpropagatL0nerrors. Nature, 323 (9) :533-536 Zheng Zhaochang. Mechanical vibration (in volumes). Beijing: Mechanical Industry Press, 1986 Zhuang table medium. Random vibration. Beijing: National Defence Industry Press, 1995Funahash [KlOntheapproximaterealizationofcontinuousmappingbyneuralnetworks. NemlNetworks, l989 {(2) :183-92BPNetworksResponseEvaluationMethodforNonstationaryStochasticExcitedMechanicalSystemLiQiangZhouJi (CADCenter, HuazhongUniversityofScience & TechnologyWuhan430074) AbstractFour'layer-BPnetworksisconstructedtoevaluatetheresponseofmechanicalsystemexcitedbynonstationarystochasticloads. ByuseofadaptingmodulationlearningalgorithmandonlyafewresultstOtrainthenetworkstresponseevaluationmappingcanbeobtainedinthesolutionofsystemdynamicalequa-tionThecomparisonresultsprovethatthismethodisapplicableandhighefficiency. Keywords: neuralnetworks; randomvibration; dynamicresponse ~ nonstationarystochastic first author Li Qiang, male, Ph.D., January 1963, l2 students. ;; 7 相关的主题文章： Abercrombie France Packaging design multimedia dis Abercrombie Fitch uk Central air conditioning syst franklin & marshall Development of circular econom |
|
