Accuracy Assurence of Calibrated Characteristics Definition for NTC-Thermistors Based on Neural Networks with Radial Basis Functions
The research is aimed to assure the calibration characteristics accuracy of the NTC-thermistors by using adaptive radial basis neural networks. To solve this problem physical and computational experiments have been conducted. As a result of it the training sample for the temperature and resistance d...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
LLC Production and Commercial Company «FAVOR, LTD»
2017-03-01
|
| Series: | Метрологія та прилади |
| Subjects: | |
| Online Access: | https://mmi-journal.org/index.php/journal/article/view/159 |
| Summary: | The research is aimed to assure the calibration characteristics accuracy of the NTC-thermistors by using adaptive radial basis neural networks. To solve this problem physical and computational experiments have been conducted. As a result of it the training sample for the temperature and resistance data have been defined and RBF-networks models have been developed that are designed for interpolation of the inverse converting function of NTC-thermistors (Table 3). While fulfilling the condition of repeatability and reproducibility of the modelling results using statistic criteria MAD, MSE, MAPE and MPЕ, it has been determined that using RBF-networks allows assuring high accuracy defining of the thermistors calibrated characteristics in the operational range of temperatures (Table 5). It has been shown that operational margin of the neural network approximation of the thermistors calibrated characteristic in lower than the acceptable error of the mathematical model, which is used in the software of modern systems of collection and processing of the measurement information (Table 9, Fig. 6). Using RBF-networks at testing software support of measurement instruments and automation of the procedure of the NTC-thermistors periodic calibration at the operational stage has been substantiated (Table 10, Fig. 10). |
|---|---|
| ISSN: | 2307-2180 2663-9564 |