Infrared Detection Technology Of Water Content In Diesel For Vehicle Application

Fuel oil（diesel oil and gasoline）is the main energy supply for heavy-duty vehicles and equipment.The water content in its components will directly affect the performance of the engine and even cause incalculable accident losses.Because of the complexity of diesel composition,if special instruments and equipment are used to detect according to the national standard method,it is not only time-consuming,but also unable to monitor online.It is difficult to meet the needs of vehicle oil safety detection.Therefore,a safe,efficient,high-precision and miniaturized vehicle diesel water content detection technology was developed in this paper.The main research contents are as follows:（1）A theoretical model of spectral absorption of aqueous diesel in the infrared band is established.The characteristic absorption peaks of double frequency and combined frequency generated by C-H,H-O and-CHx groups in the diesel sample are studied.The influence of light loss caused by turbidity on the absorption spectrum of emulsified aqueous diesel sample was analyzed by introducing the scattering theory.（2）In this paper,the infrared spectral data modeling method of watery diesel oil is given.By combining the preprocessing method with a variety of machine learning modeling regression algorithms,35 groups of models are given to predict the water content in diesel by spectral data.Due to the small storage size and low computational efficiency of in-vehicle sensor data,a competitive adaptive reweighted sampling method is introduced to select feature points for spectral data,which reduces 512-dimensional spectral data to 23-dimensional.Savitzky-Golay smoothing+Gaussian process regression model performs best.Further Bayesian optimization of the model shows that the root mean square error is 46.68×10^-6（v/v）and the average absolute error is 37.99×10^-6（v/v）,which achieves the goal of removing redundant data and improving computing efficiency.（3）A metal waveguide capillary is used to replace the traditional colorimetric dish.Integrating spectral detection,sample pool and light source,a linear gradient adjustable filter+photodetector scheme driven by motor is proposed for spectral detection.A near infrared vehicle mounted diesel sensor is developed,which can realize on-line detection of flowing samples.The experimental results show that the root mean square error is 18.99×10^-6（v/v）,the mean absolute error is 13.42×10^-6（v/v）.It has achieved a good prediction effect in the aspect of trace moisture detection in vehicle.