Multi-sensor Based Formaldehyde Detection System And Cross Interference Suppression Research

Interior decoration materials and furniture contain a mass of aldehydes.The release of aldehydes and poor indoor ventilation will cause an increase of in indoor formaldehyde concentration and affect people's health.Therefore,the detection of indoor formaldehyde concentration is particularly important.Traditional formaldehyde detection methods include spectrophotometry,gas chromatography,polarography,etc.are bulky,expensive,and difficult to operate,so they are not suitable for general users.The new formaldehyde sensor is small in size,low in cost,and easy to operate.The new-type gas sensor detection method has the advantages of small size,low cost,convenient operation,etc.But there is a problem of cross interference,so the reference value of the detection result is low.In order to solve the problem of cross-interference of formaldehyde sensor,this paper designs a formaldehyde detection system combining machine learning and multi-sensor.The system uses STM32F103 as the main control chip of formaldehyde and interference factor detection terminal to collect the response values of formaldehyde sensors,ammonia sensors,ethanol sensors and temperature and humidity sensors in the simulated environment,and recorded the actual concentration of formaldehyde at the same time in the simulated environment with a professional formaldehyde detector.The data in this paper has the characteristics of low-dimensional non-sparse.XGBoost algorithm has higher comoutational precision in processing low-dimensional datasets,and its processing speed is fast,it occupies less resources,and is suitable for mobile devices such as mobile phones.Therefore,the system uses the XGBoost algorithm to suppress the interference of the formaldehyde sensor.In order to further improve the robustness of the algorithm,the anomaly data is processed by the Local Outlier Factor algorithm.In addition,the Kernel Function is used to transform the feature space of low-dimensional non-sparse data to improve the computational accuracy of XGBoost algorithm.And then,the cross-interference suppression model is transplanted to the Android App in Java,and the sensor response value collected by the formaldehyde and interference factor detection terminal is sent to the Android App through Bluetooth.Finally,the crossinterference is calculated by combining the sensor response value and the cross-interference suppression model.The formaldehyde concentration after suppression is displayed to the user on the Android page,and the result is stored in the SQLite database.Finally,each module of the formaldehyde detection system was tested and verified.The results show that each module works normally,and the root mean square error(rmse)after cross-interference suppression of formaldehyde sensor using the method proposed in this paper is only 0.0023.Compared whith the rmse before using the Kernel is reduced by 10 times.Compared with the rmse of Neural Network is reduced by 12 times.