68-74

UDC 621.385.(043.3)
DOI: 10.15350/2306-2819.2018.1.68

APPLICATION OF REGRESSION MODELS FOR ELECTRONIC ENGINEERING PRODUCT QUALITY ASSESSMENT

N. G. Moiseev, Yu. V. Zaharov
Volga State University of Technology,
3, Lenin Square, Yoshkar-Ola, 424000, Russian Federation
E-mail: ZaharovYV@volgatech.net

ABSTRACT

Introduction. There are many approaches to solving the problem of electronic engineering product technical state prediction. Although the used mathematical apparatus is diverse, it is expedient to distinguish two main directions of the theory of prediction, differing in the product validation principle: statistical classification (pattern recognition); analytical prediction using the regression analysis method. As a result of prediction, the product belongs to one or another class with a certain quality level or a life cycle. The purpose of the work is to develop the method of the prediction of electronic engineering product quality parameters based on the pattern recognition theory using regression models. Results. Since the essence of the pattern recognition theory when solving the prediction problem consists in the assessment of the technical state of a product at the initial and final (given) moment of time, we’ll get the point value of the quality parameter using the dichotomy method (successive partition of state classes). The next stage is the construction of the mathematical model of product quality parameters dependence on factors, determining its numerical value according to m measurements during the time interval [t₀, t_m] by the regression analysis method. The mathematical model is constructed according to experimental data of products sample tests until their failure or till a given moment of time. After tests the products sample is divided into two classes: training and test samples. The model is based on the training sample material but it is checked using the test sample. The regression model adequacy is assessed according to the following criteria: residual variance, characterizing experimental data dispersion relative to results, obtained according to the constructed model; squared multiple correlation; deviation of the predicted product quality parameter value from the actual value. The obtained model is refined according to the expanded sample of products (that is more than training and test samples). The method was tested on electron-beam equipment – oscillographic tubes 11ЛО101И. Prediction models of four informative parameters of the device to the point in time 2000 hours were constructed according to the results of their control in time periods of 0, 250 and 500 hours. Conclusion. The developed technique allows assessing the technical condition of electronic engineering products according to results of their operating time at the initial stage of their performance. Its application gives the possibility of reducing life test time and predicting the quality parameter change during the given period.

KEYWORDS

prediction of quality parameters; product life; theory of pattern recognition; regression analysis

FULL TEXT (pdf)

REFERENCES

1. Fomin Ya. A. Raspoznavanie obrazov: teoriya i primeneniya; 2-e izd [Pattern Recognition: Theory and Applications; the Second Edition]. Moscow: FAZIS, 2012.
2. Lepsky A.E., Bronevich A.G. Matematicheskie metody raspoznavaniya [Mathematical Methods of Recognition]. Taganrog: Publishing house of TTISFU, 2009. 155 p.
3. Zaharov Yu. V., Moiseev N. G. Kompleksnaya ocenka kachestva izdelij ehlektronnoj tekhniki na osnove metoda statisticheskih reshenij [Comprehensive quality assessment of electronics products based on statistical decision method]. Vestnik Povolzhskogo gosudarstvennogo tekhnologicheskogo universiteta. Seriya: Radiotekhnicheskie i infokommunikatsionnye sistemy [Vestnik of Volga State University of Technology. Ser.: Radio Engineering and Infocommunication Systems]. 2015. No 1 (25). Pp. 50-54.
4. Gaskarov D.V., Golinkevich T.A., Mozgalevsky A.V. Prognozirovanie tekhnicheskogo sostoyaniya i nadezhnosti radioelektronnoy apparatury [Prediction of the Technical State and Reliability of Electronic Radio Equipment]. Moscow: Sovetskoe radio, 1974. 223 p.
5. Kartashov D.G. Forsirovannoe ispytanie apparatury [Accelerated Equipment Testing]. Moscow: Znanie, 2006. 59 p.
6. Vasiliev F.P. Chislennye metody resheniya ekstremal’nykh zadach [Numerical Methods of Solving Extremal Problems]. Moscow: Nauka, 1980. 520 p.
7. Fiacco A., McCormick G. Nelineinoe programmirovanie: metody posledovatel’noy bezuslovnoy optimizatsii [Nonlinear Programming: Sequential Unconstrained Optimization Techniques]. Moscow: Mir, 1973. 244 p.
8. Chetyrkin E.M. Statisticheskie metody prognozirovaniya [Statistical Forecasting Methods]. Moscow: Statistika, 2005. 200 p.
9. Druzhinin, G.V. Metody otsenki i prognozirovaniya kachestva [Quality Assessment and Prediction Techniques]. Moscow: Radio i svyaz', 2004. 160 p.
10. Zaharov Yu. V., Moiseev N. G. Uskorennaya ocenka tekhnicheskogo sostoyaniya ehlektronnyh sredstv [Accelerated estimation of the technical state of electronic means]. Vestnik Povolzhskogo gosudarstvennogo tekhnologicheskogo universiteta. Seriya: Radiotekhnicheskie i infokommunikatsionnye sistemy [Vestnik of Volga State University of Technology. Ser.: Radio Engineering and Infocommunication Systems]. 2014. No 4 (23). Pp. 75-81.
11. Moiseev N.G, Zaharov Yu.V. Teoriya planirovaniya i obrabotka eksperimenta [Planning Theory and Experiment Processing]. Yoshkar-Ola: VSUT, 2018. 118 p.

For citation: Moiseev N. G., Zaharov Yu. V. Application of Regression Models for Electronic Engineering Product Quality Assessment. Vestnik of Volga State University of Technology. Ser.: Radio Engineering and Infocommunication Systems. 2018. No 1 (37). Pp. 68-74. DOI: 10.15350/2306-2819.2018.1.68