Search results

The most commonly used structure is the back-propagation architecture that can work with both binary and decimal data, and can be used for a wide range of applications in comparison with other types of networks.
Then, the weights are adjusted in accordance with the reduction of the error, and the output is recalculated.
If the errors raised from the comparison are low, then the model could be taken into consideration; otherwise, the data may be inconsistent, and in this case it is necessary to modify the network architecture or the training rate.
Penetration p [mm] (experimental) Penetration p [mm] (after testing the neural network) Effective error on predicting the penetration [mm] Relative error on predicting the penetration [%] 1 2.2 2.61 0.41 1.86 2 1.6 1.69 0.09 5.62 3 2.1 2.01 - 0.09 -4.28 4 1.2 1.17 - 0.03 -2.50 5 1.8 1.56 - 0.24 -1.33 6 1.4 1.43 0.03 2.14 The MatLab source-code designed for the elaboration of the model uses the “postreg” function that determines a linear regression analyses between the target values and the output values of the trained neural network, including also the data involved in the testing phase.
The two graphs illustrate the positioning of the experimental points relative to the regression function as a result of the analysis of data considering the target ones (T) and the output ones (A) - Best Linear Fit - and in relation to the standard line to which the outputs are identical with the inputs (A = T).

(2) Data Preparation A series of Land sat TM images (1988, 1997 and 2007) were analyzed to map the main land use cover types.
The data of climate, physiognomy and vegetation maps were also used.
In this study, ERDAS IMAGINE 9.1 and ARCGIS 9.0 were used to process the data.
Socio-economic data of 21 cities and 12 counties were obtained from Liaoning Statistical Yearbook from 1988 to 2007.
It was due to the reduction of the gap area between different landscape types.

The grain growth becomes significant at high densities of ρt > 80% probably due to the reduction of pore dragging effect.
The present result is compared to the earlier SPS data reported by Bernard-Granger et al. [7].
Although the present result is almost on the same order as the earlier data, which is obtained at almost the same temperature of 1225°C, the -(σeff/Geff) relationship is inconsistent with that of their data.
Except for the difference in the densification rate, the present result is consistent well with that of the earlier data obtained by HP technique [8,16-18].

Uncertainty exists, however, in these data as it cannot be ruled out that the nuclei with new orientations might have grown from other deformed grains below the measured surface that could not be seen in the experiment.
Then the sample was cold rolled to 40% reduction in two passes.
An orientation map of the area was constructed from the EBSD data, shown in Fig.2.
All the data points with misorientation angles ranging from 41º to 39º to the nucleus are marked as grey pixels in Fig.4a.
For the present experimental data, a agreement similar to that in [3] was also observed.

These formulas established on the basis of tests and large amounts of measurement data could be accurate and reliable only if the object of research is similar to the case from which the data is obtained.
The boundary element method has the characteristic of dimensional reduction and adapting to the infinite domain, thus becomes an effective numerical method of solving the structure-borne noise.
Establish the acoustical mesh and field point mesh Establish the structure model Compute the vibration response of the structure (displacement) Acoustic transient response analysis case Define the displacement boundary condition Data transfer analysis case Acoustic mesh preprocessing set Import the acoustical mesh and field point mesh in LMS Virtual.Lab Acoustics Import the structure mesh and the transient dynamic results Set mesh parts type Define fluid material and property Fig. 3 Calculation step of vibration and radiated noise Calculation example for elevated bridge vibration and radiated noise Calculation with the moving load model.
Table 1 Maximum values of bridge vibration response Position and case Displacement[mm] Velocity[m/s] Acceleration[m/s2] Left span Model 1 3.23 0.00462 0.23137 Model 2 3.23 0.00845 1.75564 Middle span Model 1 1.64 0.00452 0.25292 Model 2 1.65 0.00875 1.92357 Right span Model 1 3.22 0.00522 0.25454 Model 2 3.23 0.00916 1.75154 Table 2 Maximum values of noise pressure[dB] Position and case Point A Point B Point C Left span Model 1 87.36 86.22 82.43 Model 2 92.63 88.64 84.87 Middle span Model 1 78.96 76.11 72.06 Model 2 89.61 83.29 79.38 Right span Model 1 92.61 87.53 85.02 Model 2 94.63 87.89 86.69 These data indicates that the displacement response based on the two kinds of vehicle model is the same, as the velocity and acceleration response has an obvious difference, the response of bridge structure with mass-spring system model bigger than the one with the moving load.

Reduction per pass was about 10%.
The grain size measurement was carried out by mean linear intercept method and the data are presented with the error bar [3] at 95 % confidence level.
The values of exponent of inverse grain size (p) at ε of 0, 0.002, 0.1, 0.2 and 0.3 were determined from the grain size and flow data based on CSR test at 220 and 450 °C, obtained from the samples undergone CR at different temperatures.
The average activation energy Q of 84 kJ Mol-1 and maximum m of 0.33 were obtained from the present data for CR of 450 °C at the test temperature of 450 °C and strain rate of 1 x 10-3 s-1, which support superplastic deformation by grain boundary sliding and its accommodation by grain boundary diffusion.
Mordike, “Magnesium Technology”, Metallurgy, design data applications, Springer Berlin Heidelberg New York, 2006

The time–frequency localization features of the wavelet transforms and their ability in improving the signal to noise ratio (SNR) make them suitable for data preprocessing [10].
Another way was estimating the standard deviation coefficient of high-frequency noise from the first layer D1 which got from the five layer wavelet decomposition as mentioned above, and then applying the soft threshold and hard threshold respectively on the signal noise reduction.
(4) Where sri is the reconstruction data in which noise have been eliminated and expressed as A5+D4+D5, soi is the original data and N is the dimension of the original data.

During the course of continuous work or driving, significant reduction in a person’s attention, concentration, sleepiness and loss of ability to sustain a required performance are some of the major concerns brought by fatigue.
The analysis system took National Instrument’s LabVIEW software as its development platform to read and analyze the EEG data.
The EEG data were saved in .txt files and processed.
Analysis results of theta data on six experiments for good and poor sleepers.
Although the research had a limited number of subjects, data showed a sign of consistency in relation to how the subject slept the night before and their current condition.

Temperature data are recorded every 10mins by using Agilent 34970A. 1- thermostatic water tank;2-valve;3-pump;4-flowmeter;5- HPETC;6- SFPC Fig. 1 The diagram of the experimental system Experimental results and analysis The entire comparative experiment is divided into two parts: one is that the performance of these two kind collectors is observed through keeping the inlet water temperature constant and changing the mass flow rate of water; the other is just the opposite.
If the experiment is performed at normal conditions so that F'(τα)e is constant and both F' and UL are constant, a straight line will result when the efficiencies are obtained from averaged data plotted against (Tm-Ta)/I according to Eq. (4).The intersection of the line with the vertical efficiency axis equals to F'(τα)e, which means that collector efficiency is at its maximum.
In addition, with the reduction in the intensity of solar radiation, heat dissipation from the SFPC to air is higher than that from the HPETC.
Fig. 3 Outlet temperature of collectors changing with different inlet temperature The data from 8:30~10:30 are chosen for the reason that the data after 11 o'clock are unstable.

This can be provided directly, for example by means of spindle driven tools, or externally by means of a drive which is not part of the machine tool. [6] The use of actively rotation tools achieves higher machining productivity at lower temperatures. [9] The temperature reduction during the machining process is achieved by rotating the tool.
From the measured data, can be seen the difference when changing the value of the feed.
Based on these data it is possible to construct a graphical representation.
Based on the performed experiment and the measured data, we can state that the lowest values ​​of the roughness parameters were achieved at a feed rate of 1.2 mm (Fig. 10).
This is evidenced by the measured data.