Design Of A Turbidity Sensor And Research On A Turbidity Prediction Method

Turbidity is a water quality parameter used to characterize the decrease in transparency due to the substances that scatter light in water.Turbidity reflects suspended substances and colloidal particles in water,and is closely related to water quality and biological activities in water.Reliable and economical turbidity sensors and the method for turbidity prediction are necessary to water quality management and process control based on turbidity changes.Therefore,the research on turbidity detection instrument and turbidity prediction methods has important research significance and application value.The turbidity sensor designed in this study based on the near-infrared scattering method has the measurement range of 0-400NTU with the characteristics of stable measurement,simple structure,and low cost.The main indicators are in line with the standard HJ/T 98-2003.The turbidity prediction method based on the multi-network fusion model proposed in this study successfully carried out 48 hours turbidity prediction in a test.The main work of this thesis is as follows:1.Research and design of turbidity sensor based on the scattering method.The optical structure design uses an infrared LED with a central wavelength of 850nm as the light source,combined with an 850nm narrow-band filter,an optical fiber,and a 3D-printed probe shell to measure turbidity by the 90°near-infrared scattering method.A STM32F303VCT6 micro control unit is applied as the main control chip of the sensor circuit and four main circuit parts including the step-down circuit,the light source driving circuit,the photoelectric conversion circuit,and the signal amplification circuit are completed.Finally,the lower computer program including DAC driver,ADC driver,light source driver,and on-chip serial port driver,is designed to complete the data measurement stably and persistently.2.Calibration experiment design and performance examination of the developed turbidity sensor.Based on the design of the turbidity sensor,the developed sensor is used to collect the data of a formazin turbidity calibration solution,and the linear model between the turbidity and the original sampling value of the turbidity sensor is established according to the experimental data,and then the model is programmed into the lower computer.Finally,the performance of the turbidity sensor is checked and analyzed through a repeatability test.The model fitting results show that the turbidity sensor has a linear relationship within the design range,and the fitting determination coefficient R~2 is 0.99.The repeatability error,zero drift,and linearity error of the turbidity sensor are 0.33%,0.15%,and 0.30%,respectively,which are in line with the standard HJ/T 98-2003.3.Research on a water turbidity prediction method.To develop a turbidity prediction method for water,1D Convolutional Neural Networks,Long-Short Term Memory,and multi-head attention mechanism are studied based on the characteristics of turbidity time series data.A multi-network fusion turbidity prediction model was developed with combining 1D CNN,Bi LSTM and Multi-head Attention.Results show that MAPE,RMSE,and MAE of the multi-network fusion model were reduced by 27.472%,8.343NTU,and 6.476NTU,respectively,and NSE was improved by 0.215 on average compared to the traditional prediction model.