«Automation and Informatization of the fuel and energy complex»

Comparative analysis of accelerometer calibration algorithms as part of unmanned aerial vehicles for monitoring oil pipelines

UDC: 531.768
DOI: 10.33285/2782-604X-2023-5(598)-45-50

Authors:

¹ National University of Oil and Gas "Gubkin University", Moscow, Russia
² National University of Science and Technology "MISIS", Moscow, Russia

Keywords: accelerometer, calibration, zero offset, axis skew, scale factor, regression model, neural network, Student criterion, confidence probability, MATLAB

Annotation:

The article considers the well-known algorithms for calibrating accelerometers based on the use of the Kalman filter, the least squares method and neural networks. Their disadvantages are discussed. The author proposed a new algorithm of accelerometers’ passport data calculation based on the construction of a regression model, where reference readings of accelerometers are used as a dependent variable, and the required factors are non-calibrated data obtained from devices when they are fixed on three axes and the rotary stand rotates in different positions. An algorithm for calculating passport data using direct distribution neural networks, each of which is responsible for obtaining a passport for a specific accelerometer is also presented; the network has three input neurons, one neuron at the output. To simplify modeling, a neural network with one hidden layer, a linear activation function of the output layer and training with a teacher is considered. The results of passport data calculations using two algorithms proved a match with an accuracy of up to the ninth decimal place or more. Two samples were compared according to the Student's criterion, one of which is the difference between the reference values and the values obtained using the regression equation, the other one is the difference between the reference values and the values calculated by the neural network model. The null hypothesis that the mathematical expectation of these differences is equal to zero was tested, contradicting an alternative hypothesis that it is not equal to zero and a confidence probability of 0,95. At the same time, the null hypothesis did not deviate, which indicates a zero difference in samples and the correctness of calibration passport calculation. This result is obtained for two algorithms.

Bibliography:

1. Balman D., Ponomarev Yu. MEMS-akselerometry kompanii Colibrys – proryv v oblast' vysokikh temperatur // Elektronika: Nauka, tekhnologiya, biznes. – 2016. – № 3(153). – S. 72–76.
2. Matveev V.V., Raspopov V.Ya. Osnovy postroeniya besplatformennykh inertsial'nykh navigatsionnykh sistem. – SPb.: Kontsern "TsNII "Elektropribor", 2009. – 280 s.
3. Veremeenko K.K., Galay I.A. Razrabotka algoritma kalibrovki inertsial'noy navigatsionnoy sistemy na dvukhosnom ispytatel'nom stende // Tr. MAI. – 2013. – № 63. – S. 10. – URL: https://trudymai.ru/upload/iblock/c8e/rus.pdf?lang=ru&issue=63
4. Vodicheva L.V., Parysheva Yu.V. Otsenka tochnostnykh parametrov datchikov besplatformennogo inertsial'nogo izmeritel'nogo bloka s pomoshch'yu otnositel'no grubogo povorotnogo stola // Giroskopiya i navigatsiya. – 2019. – T. 27, № 2(105). – S. 162–178. – DOI: 10.17285/0869-7035.2019.27.2.162-178
5. Smirnov V.A., Pravidlo M.N., Snedkov A.B. Metod sovmestnoy kalibrovki inertsial'nykh datchikov bespilotnogo letatel'nogo apparata s primeneniem neyronnykh setey // Modelirovanie, optimizatsiya i informatsionnye tekhnologii. – 2020. – T. 8, № 3(30). – S. 7. – DOI: 10.26102/2310-6018/2020.30.3.007
6. Osipova N.V. Sistemnyy analiz i teoriya prinyatiya resheniy: ucheb. – M.: Izd-vo MISiS, 2021. – 254 s.
7. Burakov M.V. Neyronnye seti i neyrokontrollery: ucheb. posobie. – SPb.: GUAP, 2013. – 284 s.
8. Nikolaeva S.G. Neyronnye seti. Realizatsiya v Matlab: ucheb. posobie. – Kazan': Kazan. gos. energet. un-t, 2015. – 92 s.