Downloaded from https://www.everyspec.com While "h or H" uniquely determines "f or F" XY are s-independent random variables if and "g or G", "1 or F" and "g or G", unique- and only if ly determining "h or H" is not true because pdf{X,Y}=pdf{X}pdf{Y} (3-7) the form of the i-dependence of x and y is not then known. The concept is the same for conditional s-independence. XY are conditionally s-inde- 3-2.4 TRANSFORMATION OF VARI- pendent random variables if and only if ABLES pdf {X, YIC} = pdf {XIC 1pdf{Y'C} (3-8) where C = a condition (event).
Let X,Y be two suitable random variables Conditional s-independence plays a very f(x) =pdf{X important role in reliability calculations j NAY) =X/f{Y where redundancy is involved. y =y(x) g(y)dy = f(x)dx (3-4a) 3-4 DISTRIBUTIONS My) =f(x) I I (3-4b) ' In reliability engineering the most com- The form of Eq. 3-4a is usually easier to re- mon domain for a random variable is (0,-). member. Variables can be transformed direct- Examples of variables with the domain (0,-) ly, within a Cdf, with no complications at all. are strength, time, failure rate.
In many cases where the domain is (--0,0), the probabilities 3-2.5 CONVOLUTION associated with (--0,0) are negligible and are IL included only to simplify the mathematics. Let This is especially true for the s-normal distri- 1. Z, X, Y be suitable random variables bution wherein negative values of some vari- with domains (- ,=) ables are physically meaningless; but it is con- 2. Z=X+ Y venient to integrate over the whole real line. 3. w(z) = pdf{Z} Continuous mathematical distributions f(x) = pdf{Y} rarely represent physical phenomena over the g(y) = pdf{Y} entire domain of the variable.
Usually, how- h(x,y) =pdf{X,Y} ever, the probabilities associated with the dis- 4 Then, the convolution formula is turbing part of the domain are negligible. If they are not, then of course, the model must w(z) = h(z-y,y)dy (xz-xkdx -) be reformulated. ~fi"4 =fh(i (3-5) = h(xz - x)dx 3-4.1 MOMENTS Random variables with pdfs have mo- If X and Y are s-independent, then the con- ments. The two conventional points about f volution formula is which to take moments are the origin and the mean; when taken about the mean, they are w(z) =ffx)g(z-x)du=ffz-y)g(y)dy called central moments. Two random vari- J -f ables can have joint moments, although only - y)g(y)dy (3-6) the second is used practically.
LeL X be the random variable and f(.) = pdf {X ). The nth moment (about the origin) of X is the s-exrpected value of x" : 3-3 s-INDEPENDENCE AND CONDITION- ALs-INDEPENDENCE f The notion f s.independence is analogous to that for discrete distributions. where X implies the integral over the domain 3-3