Search results

To solve the linearization problem of QTTS, the traditional method of using the calibration data is to perform a third- or fourth-order polynomial curve-fit by means of a least-squares regression technique, or use of a multidimensional look-up table[7].
The calibrating data is obtained by a calibration unit that is necessary for the training and the testing of the ANN.
(a) Geometry and vibration mode of tuning fork (b) Crystallography of a ZYtw-cut tuning fork Fig. 1 Quartz tuning fork temperature sensor block diagram Fig. 2 Measurement Instrument black diagram In the operation phase, the output signal of QTTS is generated and optimized by OC which its output is connected to HP5334A and transferred to PC via IEEE-488 where data reduction and optimization is implemented.
The PC-based instrument is controlled by the software written in both operation and calibration phases, which is used to acquire the data from ANN training phase and to show the calculated temperature in operation phase.
Therefore, the other data set of 50 QTTS temperatures that is uniformly distributed between -200 and 200 C is used in the test process.

The results were compared with the experimental data [1-3].
This process is considered by the model through the reduction of carbon deposit intensity beside the actual carbon flux
The factors are selected to obtain the best fit with the experimental data (see below).
The cluster surface concentration temperature dependence obtained from the computer simulation was compared with the experimental data.
The result obtained within this model appeared to be in good agreement with the experimental data.

Introduction With the development of Internet and related technology growth quite rapidity, there emerge quite large amount of Web information resources, and the amount of data is quite large and increased quickly.
SVM Decision Tree A support vector machine (SVM) is a concept in statistics and computer science for a set of related supervised learning methods that analyze data and recognize patterns, used for classification and regression analysis.
The standard SVM takes a set of input data and predicts, for each given input, which of two possible classes forms the input, making the SVM a non-probabilistic binary linear classifier.
Common methods for such reduction include: Building binary classifiers which distinguish between (i) one of the labels and the rest (one-versus-all) or (ii) between every pair of classes (one-versus-one).
When the SVM decision tree algorithm was used for data classification, to transform into various forms, the structures of corresponding binary tree are also varies.

However, the need exists since, for example, investigations on noise reduction in gear units is an essential field of research in the automotive industry.
Unfortunately, it almost always leaves out of consideration that the utilisability of FFT algorithms is bound to strict preconditions and that the spectra which are determined by applying the FFT to measurement data are, by no means, always identical to the really existing spectra.
The application of the FFT algorithm to the measured gear data yields the spectrum shown in Figure 4.
The spectrum determined is the correct FFT spectrum which must be attributed to the corresponding data.
The method applied to the data is, under the given conditions, not suitable for determining the actual spectrum. 0,000 0,002 0,004 0,006 0,008 0,010 0,012 0 10 20 30 40 50 wavelength in mm amplitude in mm 0,000 0,002 0,004 0,006 0,008 0,010 0,012 0 10 20 30 40 50 wavelength in mm amplitude in mm Figure 4 Spectral assessment by means of a) FFT algorithm b) Bayes' statistics In the case of these measurement data, the conditions stated for the application of the FFT for the waviness analysis were not fulfilled so that another assessment method had to be used.

With the rapid increase of image data, how to retrieve has become more and more important.
Lots of image data are stored and transmitted in compressed format, but the study of image retrieval technologies are mostly not based on the data of image compression at present.
When we retrieve the compressed image data, we need the “full decompression operation”.
They. transformed DCT domain coefficients for the Mandala domain data.
Because when the ratio increases to a certain degree, the coefficient in retrieval behind process will lose the function, which is equivalent to the reduction of DCT coefficients in the number when retrieving, to some extent, will weaken the descriptions of the image information.

The working principle is as follows: first the SCM system sends commands to the driver of stepper motor, second the driver chip drive stepper motor, third the robot move to the joint through reduction gear, then the peristaltic pump starts to dispense, At last the robot returns to the initial position.
The upper microcontroller is responsible for the management of entire system, the calculation of kinematics, the planning of trajectory, detecting the signal of sensor, sending data to the lower microcontroller.
The lower microcomputer is mainly responsible for receiving the data, and then controlling stepping motor to drive manipulator to move at the accurate positioning and controlling peristaltic pump to finish gluing task.
Fig. 6 Hardware Structure Diagram The upper and lower machine both adopts AT89C52 in the control system the manipulator, the upper machine is shouldering the task of processing data, and ATM89C52 only contains 256bytes random access data storage, so it is necessary to extend a piece of static data memory RAM6264 to meet the requirements; the upper machine not only can successfully completes the communication with the lower machine, but also realizes the communication with the PC through the switch; when the lower machine receives the value of each joint angle changes, it will control STK672-04 to drive the stepper motor, can achieve the manipulator motion; when the lower machine receives the signal to send dispensing or stop signal, it will control the peristaltic pump to finish dispensing.

The establishment of the finite element model Take type of XB1-120-100 harmonic gear drive for example, the reduction ratio i=100, the modulus m=0.6mm and the pressure angle α=20°.
Recording the value of input torque Tin corresponding to the value of output torque Tout, whereby the efficiency curve shown in Fig.2, the experimental data in Table 1. 2) In the case of maintaining the output torque (Tout=100N·m), the input rotational speed nin is increased until the value up to the rated speed.
Recording the value of the input torque Tin corresponding to the value of input rotational speed , whereby the efficiency curve shown in Fig.3, the experimental data in Table 2.
Table 1 The experimental data of harmonic gear drive efficiency when the input rotational speed (nin =3000rpm) unchanged nin(rpm) nout(rpm) Tin(N·m) Tout(N·m) Pin (kW) Pout(kW) μ(%) 3000 30 0.4 8.21 0.125 0.025 20.4 3000 30 1.26 90.76 0.398 0.285 71.5 3000 30 1.57 118.11 0.49 0.37 74.8 3000 30 2.21 177.69 0.697 0.558 80.0 3000 30 2.97 246.63 0.933 0.774 83.0 3000 30 3.58 305.67 1.125 0.960 85.3 3000 30 4.22 364.15 1.326 1.143 86.2 3000 30 4.75 414.08 1.490 1.300 87.0 3000 30 5.14 448.27 1.610 1.615 87.1 Fig.2 The curve of the relationship between output torque and efficiency when the input rotational speed is nin =3000rpm The data obtained from the Table 1 shows that, when the input rotational speed unchanged, with the output torque increase the efficiency of the harmonic gear drive increased.
Table 2 The experimental data of harmonic gear drive efficiency when the output torque (Tout=100N·m) unchanged nin(rpm) nout(rpm) Tin(N·m) Tout(N·m) Pin (kW) Pout(kW) μ(%) 68.4 0.68 1.51 100 0.476 0.314 65.8 968.4 9.68 1.36 100 0.429 0.314 73.0 1256.4 12.5 1.36 100 0.429 0.314 73.1 1552.8 15.5 1.39 100 0.437 0.314 71.7 1741.2 17.4 1.36 100 0.430 0.314 73.0 2071.2 20.7 1.40 100 0.440 0.314 71.3 2370 23.7 1.38 100 0.433 0.314 72.4 2679.6 26.7 1.38 100 0.433 0.314 72.4 2955.6 29.5 1.36 100 0.429 0.314 73.1 Fig.3 The curve of the relationship between input rotational speed and efficiency when the output torque is Tout=100N·m The date from the Table 2 shows that when the output torque unchanged, with the input and output rotational speed increasing, the input and output power increased, therefore the transmission efficiency of harmonic gear drive almost no change.

In this paper, experimental research has been made about parabolic trough solar water heater,which provided basic data for design and installation of water heaters and had a major push and reference to the development of solar collector.
The time interval between the adjacent test is set to 30min.Instantaneous efficiency of different moments in every day can be calculated by the test data and Formula (1).Figure2 shows the change in temperature of water in tank during the day .
As can be seen from Figure3, Figure4and Figure5,the instantaneous efficiency was changing with time,but on the whole,the instantaneous efficiency increased first and then decreased,the instantaneous efficiency reached the maximum at about 12:00.Meanwhile,during the time of 9:00am to 3:00 pm,the instantaneous efficiency was relatively stable.The unsteady equation of the solar hot water heaters can be concluded with the method of least squares by data obtained in several tests [5]
The heat loss coefficient was calculated by Formula (5), the measurement data is shown in table 3
（5） Where:—Capacity of water tank ,kg;—Heat capacity of water,; —initial temperature at night, ˚C; —end temperature at night, ˚C; A — volume of water tank,L; —average temperature in the water tank at night, ˚C; —average temperature of the around environment at night, ˚C;—test interval, s Tab.3 The heat loss coefficient measurement data table Date Initial temperature of the water (˚C) Final temperature of the water (˚C) Average temperature of the water (˚C) Ambient temperature (˚C) (W/(m2•˚C)) 2013.4.7 52.9 48.8 50.85 8.1 1.017674 2013.4.18 35.8 33.1 34.45 7.6 1.067039 Conclusions (1)The overall trend of the instantaneous average efficiency changes of the parabolic trough solar water heater is firstly and then decreased.

In response to this shortcoming, many scholars have used a method of combination with PCA and LDA [8], the advantages of the PCA and LDA together fully integration, and it can not only solve the problem of PCA algorithm is not sensitive to the different training sample data problem, but also LDA algorithm when the within-class scatter matrix is singular, and obtain a better classification results.
The sampling time for each sample is 60 seconds and the rinsing time is set as 110 seconds, The interval for data collection was one second.
When the measurement was completed, the obtained data was stored in a computer for later analysis.
The reason is when the difference of sample quality grade is small, there is a big overlap of information or relevance in the differences in the sample that reflect by electronic nose sensor , PCA algorithm to find only the data distribution of spindle orientation, retained after dimensionality reduction by the information is not necessarily the most effective for classification .
Conclusions In this paper, PEN3 electronic nose was used to test Atractylodes samples of three growing areas, data analysis method using LDA algorithm based on MSD criterion to solve the problem of small samples, also distinguish with three Atractylodes from three different growing areas correctly, and the correct recognition rate of all testing samples reaches 97.8%.Furthermore, the classification results clearly superior to the use of PCA or PCA + LDA algorithm.

PPCA model [2-4] is a generated model by adding the noise with the assumption that its variance is isotropic to Factor In fact, PPCA model is also a latent variable model with factor analysis method and its mathematical model is given by the following equation (1) where X is the original variable and represents a matrix sample data with normalized processing, m is the number of samples, and n is the number of original variables, P is a load vector not known previously; e (t) is Gaussian noise.
After the original data X is decomposed into the summary of the outer product of load vector P, the main element model can be expressed by (2) where represents the main element column vector, is the column vector of load.
From Fig. (2), we can see that each main element is the projection of original data X onto the direction of the load, which are selected to reconstruct new data, ultimately a reduction matrix with lower dimension is got and the de-noising purposes.
The test system consists of speed monitor, manual speed governor, acceleration sensors, speed sensors, motors, spindles and computer with VQ data acquisition software.
In the experimental study, the sampling frequency is 25.6 kHz and 327680 data points are collected.