- •Content
- •4.1 Introduction 40
- •5.4 Conclusion 71
- •Abbreviations
- •Introduction
- •1. Experimental research methods
- •1.1. Planning an experiment in research
- •1.2. Measurement error
- •1.4 Errors of indirect measurements
- •1.3 Processing and analysis of experimental data
- •1.5 Methods of the theory of correlation
- •1.6 Dimensionless criteria of the atomization process
- •2. Experimental instrumentation
- •2.1 The common-rail fuel injection equipment
- •2.2 Review of drop sizing techniques and spray characteristics
- •2.3 Direct imaging method
- •2.4 Formation of liquid spray
- •2.5 Conclusion
- •3. Analysis of mixture formation and flame development of diesel combustion
- •3.1 Experiment set up
- •3.1.1 Rapid Compression Machine
- •3.1.2 Optical Setup
- •3.1.3 Direct Photography
- •3.1.4 Schlieren Photography
- •3.2 Summary
- •4. Analisis of spray development and autoignition process
- •4.1 Introduction
- •4.2 Theoretical Analyses
- •4.3 Experimental set-up
- •4.3.1 Constant Volume Combustion Chamber (cvcc)
- •4.3.2 Camera system
- •4.3.3 Nozzles and theirs dimension
- •4.3.4 Fuels
- •4.4 Experimentl technique
- •5. Experimental results
- •5.1 Spray tip measurments
- •5.2 Cone angle measurments
- •5.3 Autoignition Measurements
- •5.4 Conclusion
- •6. Безпека I охорона в морi
- •6.1 Підготовка членів екіпажу з охорони судна: планування, організація занять і навчань
- •6.2 Застосування плану охорони судна
- •6.3 Основи захисту моряків в умовах хімічного забруднення судна
- •8. Охорона працi
- •8.1 Стомлення і перевтома
- •8.2 Професійний стрес
- •8.3 Теорія горіння, пожежний трикутник
- •8.4 Знезараження та очищення стічних вод
- •9. Техніко-економічне обгрунтування введення досконаліших систем контролю робочих процесів суднових дизелів
1.5 Methods of the theory of correlation
Apply when the two values x and y are not linked func–tional dependence, but are probabilistic relationship. This relationship is ap–lyaetsya usually when there is general random factors affecting as one and to another value, along with other unequal values for both random factors, such as,
;
.
With such a relationship between x and y makes sense to talk about the law of distribution of these variables on the probability with which there are certain combinations of values. To do this, consider the function that creates the expectation of a size at different fixed values of the other. These functions are called regression, and their graphic representation in the plane (x, y) (regression line) is the locus of the centers of conditional distributions (center of mass), corresponding to a given value of one variable.
Along the lines of the regression line review and conditional variances, which is a measure of dispersion of a value at fixed values of the other.
By analyzing these relationships allows to solve the problem of sound prediction, ie to find the extent to which reliability with preset contains the value we are interested if other related quantities are of some importance.
The most simple and important cases of regression lines are straight lines and the corresponding regression equation in x in x and for y. The characteristic connection between the values of x and y in this case is the moment of connection μху, a measure of last – dimensionless coefficient called correlation coefficient:
where
— standard deviation.
Numerical linear correlation coefficient can be determined by the formula:
where
;
;
n – number of
trials.
The basic properties of linear correlation coefficient lies in the fact that its value is within the ± 1 that it is only when between x and y is exact and thus, a linear relationship. If | r | <1, then it loses these properties. However, as we approach | r | to one, with some reason it still considered a measure close to full linear relationship between x and y. Where r = 0 say that x and y are not correlated and, therefore, independent.
If | r | is very different from unity (and especially when r = 0) value of x and y is expedient to investigate nonlinear correlation dependence. To find the latest use of computers through which selected the most suitable approximated equation that gives the greatest approach to regression curve.
In practice, the research is necessary to establish the links between the three and a large number of physical quantities suppose z, x and y, depending on the total and varying them random factors. Distress communications in this case the estimate cumulative or multiple, correlation coefficient:
Coefficient
can
take values: 0<
<1.
When
between z and x and у there is no linear relationship with = 1 value z is a linear functional dependence on x and y, as expressed, for example, the equation:
Effect size x (or y) to replace r evaluate partial correlation coefficient.
Similarly, define r (z, y). Some correlation coefficients have the same properties as the linear correlation coefficients.
