A.B. Kuzmin, N.D. Prigonyuk, A.V. Nikonenko

Peculiarities of investigating statistical dependence between controlled parameters are discussed. The main attention is paid to the use of obtained relations for decreasing the number of failures of a radioengineering object at the expense of tolerance corrections of the controlled parameters. At the presence of N similar radioengineering objects performing the works connected with control of the given parameters, the number of refused objects is determined and the probability of failure of an object is evaluated for determination and control of a technical state of each object. Failure probability estimates are also computed for each quantitative and qualitative parameter. The technique is proposed for defining the statistical dependence (or independence) between parameters and revealing the type and degree of such dependence. To do this, in the Cartesian coordinate system, the scattering diagram is plotted which allows qualitative evaluation of the type and degree of dependence between parameters. Mutual positions of points give information about qualitative dependence (mean, strong, positive or negative) or independence of parameters. Approximation of points position with using the determinate function @ is performed for dependent parameters. The correlation coefficient @ serves as a qualitative estimate of the type and degree of linear dependence between parameters. If the MSE value at approximation satisfies a customer exploiting an object, than the dimension of the controlled parameters n can be reduced and thus the material costs on object exploitation also can be decreased. For example, let one parameter be controlled for a long time with the use of an expensive device while the other parameter be controlled for a short time period with a cheap device. Here one should control the second parameter and evaluate the first parameter value. The developed technique has a practical value at exploiting radioengineering objects and allows one to reduce material costs at the exploitation with decreasing the number of objects failures at the expense of tolerance corrections of their parameters.