: pp. 37-41
Lviv Polytechnic National University, Ukraine
Lviv Polytechnic National University

Current article considers the results of the study of air and water flow temperature prediction error on the number of  inputs  in  neural  network.  Authors  guide  the  architecture  of  neural  network  for  temperature  prediction.  The  formula  of temperature step  response  for  real sensor  is given. Also,  the method  for  calculating  the  time  constants  for  the  temperature step response formula using real measurement data is considered.

The purpose of the present article is to study the dependence of the temperature prediction error on the number of inputs in neural network and to verify the neural network on real measurement data. First, an algorithm for creating and calculating the test sequences for neural network training is researched. Second, the neural network training this algorithm is studied for predicting the temperature on the basis of the real measurement data. The last are received in the laboratory of Institute of Process Measurement and Sensor Technology in  Ilmenau University of Technology. The test equipment for air and water temperature measurements is described. The measurement of air  temperature was performed with J6-type  thermocouple. Air  temperature measurements were performed at the air velocity 1.0, 3.0 and 5.0 m/s. For each case the temperature predictions have been fulfilled by means of the neural network with 10, 20 and 40 inputs. The table with maximal air temperature prediction error is given.

The measurement of water  temperature was performed with N-type  thermocouple  at  the water velocity  of 0,2 m/s. For prediction, neural networks with 20 and 40 inputs have been used. For both cases, the prediction error was accessed as practically the same.

Conclusions about air and water temperature prediction results have confirmed that the computed results coincide with the experiments.

[1]  F.  Bernhard,  Handbuch  der  technischen  tempera-turmessung. Springer Vieweg, 2014. 

[2]  N.  Yaryshev,  Theoretical  basis  for  measuring  non-stationary temperature. Leningrad, USSR: Energoatomizdat, 1990.  

[3] L. Michalski, K. Eckersdorf, J. Kucharski, J. McGhee, Temperature measurement, John Wiley & Sons, Ltd, 2001.

[4]  N.  Kovalchuk,  E.  Polischuk,  I.  Pytel,  K.  Semenistyi, “Modern  methods  and  means  of  determining  the  dynamic characteristics of converters”, iss.1, TS-6, NIITEI Instrumentation, 1983. 

[5] D.  Kriesel, A  Brief  Introduction  to Neural Networks, 2007,  [Online].  Available:     http://www.dkriesel.com/en/science/ neural_networks

[6]  R.  Bordawekar,  B.  Blainey,  R.  Puri,  Analyzing Analytics. Morgan & Claypool Publishers, 2015. 

[7] O. Lopatko, I. Mykytyn, “Temperature value prediction errors  using  neural  networks  and  ideal  transition  process”, Measuring equipment and metrology, no.78, p.20-24, 2017.

[8]  S.  Augustin,  T.  Fröhlich, M.  Schalles,  S.  Krummeck, “Bilateral comparison for determining  the dynamic characteristic values of contact thermometers in fluids”, Journ. sensors & sensor syst.,  2018.  [Online]. Available:        https://doi.org/10.5194/jsss-7-331-2018.

[9] F. Lieneweg, Übergangsfunktion (Anzeigeverzögerung) von  Thermometern  –  Aufnahmetechnik,  Meßergebnisse, Auswertungen,  Archiv für  Technisches  Messen,  1964,  R46–R53, 1964.

[10] H. Mammen, G.  Krapf, C. Hoffmann,  T.  Sasiuk, M. Pufke,  S.Marin,  T.  Fröhlich,  “Prüfeinrichtung  zur Untersuchung des  dynamischen  Verhaltens  von  Berührungsthermometern  in Wasser”, in Proc. TEMPERATUR 2017, 17 and 18 Mai 2017, PTB Berlin, Tagungsband, 2017, pp.163–168.