Thermosensitive thermometer calibration method based on neural network

Abstract

The invention provides a thermosensitive thermometer calibration method based on a neural network. The method comprises the following steps: 1) obtaining sample data value, comprising measured temperature in a certain section and a corresponding thermistor value; 2) carrying out neural network model arrangement; 3) training a thermosensitive thermometer calibration compensation model; and 4) correcting the measured result of a thermosensitive thermometer by utilizing the obtained calibration compensation model, wherein the step 3) also comprises the following steps: step31) initialization of network parameters, step 32) initialization of the neural network, step 33) calculation of network errors and step 34) network training. The method fully considers nonlinear characteristics of measurement of the thermosensitive thermometer, constructs the calibration compensation model by utilizing the feature that the neural network can approximate to a nonlinear function better, and can improve measurement accuracy of the thermosensitive thermometer; and besides, the method does not need to additionally increase a peripheral hardware circuit, does not increase test cost, and has a certain practicability.

Description

technical field [0001] The invention belongs to the technical field of measurement, and relates to a heat-sensitive thermometer, in particular to a neural network-based calibration method for a heat-sensitive thermometer. Background technique [0002] At present, the measurement principle of general thermometers is that the temperature sensor converts the temperature into electricity, and the electricity is converted into a digital temperature value through an electronic circuit. Due to the nonlinear characteristics of the electronic circuit, it will inevitably cause the nonlinearity of the thermometer, so that the actual temperature cannot be accurately measured, resulting in a system error; in addition, additional errors will also be formed when the environmental variables such as the temperature measurement distance and the temperature measurement delay change, especially when the temperature measurement method is used. When measuring temperature with a sensitive thermome...

AI Technical Summary

Problems solved by technology

Due to the nonlinear characteristics of the electronic circuit, it will inevitably cause the nonlinearity of the thermometer, so that the actual temperature cannot be accurately measured, resulting in a system error; in addition, additional errors will also be formed when the environmental variables such as the temperature measurement distance and the temperature measurement delay change, especially when the temperature measurement method is used. When measuring temperature with a sensitive thermometer, there are differences between each specific thermistor, and there is also nonlinearity between temperature and resistance, so thermometer calibration is very important

In order to improve the measurement accuracy of thermometers, most manufacturers adopt the method of hardware improvement, by compensating and calibrating the resistance value of the thermal temperature sensor, so as to improve the measurement accuracy of temperature. The disadvantage is that it needs to increase the peripheral circuit, which increases the test cost; There are also a small number of manufacturers adopting the method of software improvement, using the least square method to simulate the simple temperature linear characteristic curve, and then performing temperature compensation. Because it does not meet the nonlinear characteristics of thermometer measurement, the calibration results are not ideal.

Images