The Design And Implementation Of A Sensor Data Compensation Algorithm

The accuracy of the sensor plays a very important part in decteing system, but errors between actual value and measured value always exist because of the influence of surroundings. It has been a hot issue how to put lineal or non-lineal error compensation into effect.Various methods of error compensation are compared in advantage and disadvantage, and an improved new one based on improved Partial Least Squres (PLS) is put forward in this paper. The new method fit a realational expression between correction factor (the ratio between actual value and measured value) and the surroundings, it introduces high order of envirmental parameter as new variables, some of which are excluded by correlation analysis. After resolving the model, the polynomial about envirmental parameter is getted, and compensation to the sensor is achieved. In actual measurement, correction factor could be calculated according to current envirmental parameter, then actual value is obtained. Error coule be reduced to a reasonable data through adding high order polynomial of evirmental parameter. The new method could be applied to different error, it improves the disadvantage of the traditional which has a simple model.The new compensation method is used in the module of data collection and pretreatment in overall security evaluation system.Temperature and humidity are chosen as the parameter in compensation, and the last regession relation about the correction factor is composed by high order polynomial of temperature and humidity. The test result shows that the compensation method based on improved PLS has a big progress in accuracy, compared with the one based traditional PLS.