Search

The KLASS computer program planned for development is designed to work as a part of the ARION-plus software package and allows you to perform an automated assessment of the reliability of electronic products, including semiconductor devices. At the stage of work planning on the creation of the KLASS program, as a module of the ARION-plus complex, the question arose about the working time allotted for the procedure for testing a computer program. The approaches described in the scientific literature used to assess the operational reliability of computer programs, taking into account their testing, proceed from the fact that the program code has been written and debugged and there are certain data on the results of testing the computer program. Software developers would like to know the predicted testing time, which ensures a given operational reliability of a computer program, even before starting work on writing program code. Based on the analysis of the experimental data on the reliability of computer programs in various fields of application, a model is proposed for determining the testing time required to ensure the operational reliability of programs. The model was used for the computer program KLASS planned for development and takes into account the programming language, the amount of program code, the speed of the computer processor, and the scope of the program. Based on the obtained model, a nomogram with two binary fields was constructed, which allows one to quickly determine the predicted time for testing computer programs.

When evaluating the individual reliability of semiconductor devices by gradual failures for a given operating time, the value of the electrical parameter of a particular instance for this operating time is predicted using the simulation method. To obtain a prediction of an electrical parameter, it is necessary to have a simulation model in the form of a function of the relationship between a given operating time and the level of simulation impact. The simulation model is obtained with the help of preliminary studies (training experiment) of a training sample of semiconductor devices of the type of interest with the volume of approximately 50 ... 100 copies. The application of the model is reduced to the calculation of the level of simulation impact corresponding to a given operating time. The result of measuring the electrical parameter at the calculated level of the imitation impact in a new specimen of the same type that did not take part in the training experiment should be considered as a prediction of this parameter for a given operating time. An integral part of the preliminary research to obtain a simulation model is the testing of semiconductor devices of the training set for a long operating time, which can be tens of thousands of hours, which necessitates planning and conducting accelerated tests. The article substantiates the conditions for conducting accelerated forced tests in relation to high-power bipolar transistors of the KT872A type. The increased temperature and the reverse voltage applied to the collector junction of the transistors are chosen as factors accelerating the tests. The test acceleration coefficient is calculated relative to the operating mode of the transistors. Based on the results of the accelerated tests for the electrical parameter (collector-emitter saturation voltage), a mathematical model was obtained in the form of a dependence of its average value on the operating time. The presence of this model is necessary to determine the function of recalculating the given operating time of transistors to the value of the simulation impact.

The threshold logic method used for electronic products when assessing their reliability in the form of predicting the performance class of products for a given operating time (K1 is a class of operable ones, K0 is a class of inoperable copies) provides for the transformation into a binary code of informative parameters obtained at the initial moment of time, and allows to build a forecasting model in the form of a table showing which code combinations correspond to the instances of the K1 class. The use of a binary transformation simplifies the forecasting procedure, but the reliability of the predictions is slightly reduced. It is relevant to obtain a forecasting model that has the simplicity of its application and provides a higher reliability of forecasting a class of products than with a binary transformation of parameters. On the example of bipolar transistors of the KT872A type, the regularities of electrical parameters used as informative ones are established, and for their transformation into a code, it is proposed to consider three areas of parameter change: the first area is the range of values between the mathematical expectations obtained separately for instances of the classes K1 and K0; the other two areas are the parameter values to the left and right of this range. Parameter values falling within the range are assigned the R code (from the word Range), outside the specified range - the code 1 (one) or 0 (zero), depending on the pattern of the informative parameter. It is explained how to convert parameters to codes 1, 0 and R and get a prediction model in the form of a logical table built from these codes. Basing on the example of the investigated transistors, it is shown that the proposed heuristic model provides the best forecasting results, practically keeping the simplicity of the basic method of threshold logic.

The reliability of computer-based information systems is largely determined by the reliability of the developed application software. The failure rate of its computer program is considered as an indicator of the reliability of the application software. To determine the expected reliability of the application software planned for the development (until writing the code of a program), the model is proposed that uses some parameters of the future computer program, data on the influence of various factors on its reliability, and further testing of the program. The model takes into account the field of software application and computer processor performance. The process of model parameters obtaining is analyzed., It is possible by use of proposed model to determine the predicted failure rate of the planned application computer program, and then the reliability of the computer-based information system as a whole. If necessary, the measures can be developed to ensure the required level of reliability of the computer-based information system.

When assembling electronic complexes for medical purposes, it is important to install highly reliable semiconductor devices in electronic equipment. Experimental studies and the example of high-power bipolar transistors in this work show how you can select copies of an increased level of reliability for their subsequent installation in critical electronic devices. To select highly reliable samples, individual forecasting was used according to informative parameters measured for a particular sample at the initial moment in time. Experimental studies (training experiment) included measuring at the initial moment of time for each sample of transistors of electrical parameters, which may contain information on reliability, and then conducting accelerated tests of transistors for reliability for a time corresponding to normal operating conditions specified in the technical documentation. The training experiment is performed once and used to obtain a predictive rule, which is applied to other similar samples that did not participate in the training experiment. To obtain a predictive rule, the method of majority logic was used. Prediction is performed in the form of assigning a specific sample to the class of highly reliable samples for a given future operating time. To perform prediction, the values of the informative parameters are measured at the initial moment of time for a particular sample of interest, they are converted into binary numbers (zero or one) using the threshold values found from the results of the training experiment, and the decision on the correspondence of the sample to the class of highly reliable transistors is made by a set of binary numbers. To classify a sample as a highly reliable one, it is sufficient that the number of ones exceeds the number of zeros in the resulting set of binary numbers.

Individual forecasting of the reliability of semiconductor devices, taking into account gradual failures, is an urgent task, as it allows you to choose highly reliable instances for critical electronic devices of long-term functioning. In relation to bipolar transistors, an approach is proposed that allows us to solve this problem by using the voltage applied to the collector-emitter junction as a simulated effect. Using the example of highpower bipolar transistors of the KT872A type, it is shown how the problem is solved. For the sample of transistors of this type using the results of a training experiment, two equations were obtained to describe the electrical parameter under consideration (a static base current transfer coefficient in a circuit with a common emitter), the value of which judges the absence or presence of a gradual failure for a specific instance. The first equation shows how the electrical parameter changes on average depending on the voltage applied to the collector – emitter junction. The second equation describes the average degradation of the electrical parameter during long-term operating time of transistors. Based on these two equations, a simulation model of the reliability of bipolar transistors of the type in question is obtained in the form of a communication function that shows what level of simulation voltage corresponds to a given operating time. As applied to the transistors of the type under consideration, the obtained simulation model allows us to individually predict reliability by the gradual failures of the same type of samples that did not participate in the training experiment. To do this, first determine the value of the simulation voltage corresponding to a given operating time. This is achieved by substituting a given operating time into the model. The individual forecasting of the reliability of a new onetype instance consists in measuring the electrical parameter of this instance at a voltage on the transistor collector corresponding to the calculated simulation value, and comparing the measurement result with the norm set on the electrical parameter.

To predict the expected reliability of the developed software applications, a method based on the use of the statistical model of the operability of software is proposed. The model is based on generalized statistical data taken from domestic and foreign printing on the expected number of errors in the developed application software, depending on the number of lines of computer program code, the duration of its testing, the qualifications and experience of testers. An example of predicting the expected reliability of an application software is given

The method of obtaining the model, describing degradation regularity for function parameter of electronic device (ED) sample, was systematized. The probability distribution function (for operating time of interest) of function parameter of ED sample is considered as a model of degradation. On the basis of the method the technique for prediction of parametric reliability of new ED samples was developed. It can be used for new samples of ED of the same type, but which did not participate in preliminary research/investigations. Prediction is obtained as the probability of the fact, when functional parameter of any device in a sample will have the value within the specified limits after specified operating time.

A new approach to the evaluation of the functioning of the electronic systems for medical purposes is proposed. This approach takes into account the sustainable system failures and possible temporary system device failures due to the impact of their work of natural and industrial influences.

For crystals of EEPROM integrated circuits (ICs) a method for predicting of information storage time after the power is turned off is provided. Prediction is performed using the accelerated tests, which are considered as the temperature effects that accompany the technological operations in the premises of the crystal body and the ICs assembly. All technological IC assembly operations have complex effect. For this effect, estimated coefficient of acceleration tests is founded. Also for normal operating conditions of the ICs it's found the guaranteed time of information storage after power off.

With the help of accelerated tests we have obtained experimental data about the degradation of the functional parameters of three types of high-power transistors as the representatives of electronic products. Histograms of parameters’ distribution were constructed on the experimental data. It was found that the normal distribution law is deformed during the operation time and it poorly describe the parameters degradation for samples of products. The results of the degradation of the functional parameters are used for solving problems of group and individual forecasting of reliability of transistors taking into account the gradual failures.

The authors propose a method for estimating the expected reliability of applied computer programs for medical computer systems in the early stages of their design. To obtain the expected operational reliability of a computer program, the method uses correction factors that take into account the most important factors determining the initial level of program reliability (before testing begins). Accounting for the impact on the reliability of the program testing procedures are performed using a correction factor that depends on metrics that take into account testing technology, testing time, experience and qualifications of specialists.

For semiconductor devices of high power, the authors experimentally obtained of models of functional degradation parameter in the form of the conditional density of its distribution under the assumption of three hypotheses about the distribution law at the points of operating time: normal, two-parameter exponential and Weibull-Gnedenko. Using the average prediction error, the authors compared the effectiveness of degradation models in determining of the parametric reliability of new samples of semiconductor devices of the same type.