Research And Design Of Hydrogen Sensor Detection System Based On Improved BP Neural Network

As a highly efficient and clean secondary energy,hydrogen has been widely used in military,power generation and automobile.Due to the inflammable and explosive characteristics of hydrogen,a large number of hydrogen sensors should be used for realtime detection in practical applications to avoid leakage and explosion accidents.Although various types of hydrogen sensors have been prepared based on different measurement principles,due to the limitations of the physical properties of gassensitive materials,they will be more or less affected by external environmental factors,resulting in the decline of system stability.To solve these problems,an intelligent hydrogen sensor detection system considering environmental factors is proposed in this dissertation.The upper computer software embedded with genetic algorithm to improve BP neural network was built to collect the output electromotive force and various environmental parameters of the sensor through the lower computer.After training and learning of the optimized model,the hydrogen concentration under the corresponding environment was accurately output to improve the anti-interference ability of the sensor detection system.First of all,for semiconductor and thermal conductivity hydrogen sensors that are greatly affected by environmental interference,through the research on their detection mechanism,detection circuit,output electrical signal,etc.,as well as the sensitivity,temperature and humidity characteristics of semiconductor sensors,the influence of temperature,humidity and gas concentration on the sensitivity of the sensor is obtained.Secondly,this dissertation designs a detection system based on thermal conductivity hydrogen sensor,and provides a design plan for peripheral hardware conversion circuits such as signal amplification and filter,A/D converter,RS485 communication circuit,circuit power supply and microcontroller circuit.Clarified the specific process of the lower computer controller program design and the system upper computer design,and completed the lower computer PCB board circuit design,welding,functional testing and the upper computer software program development,communication and debugging.Thirdly,on the basis of the data collected by the sensor detection system,the BP neural network algorithm is improved by using genetic algorithm,and the network models of the output response of semiconductor and thermal conductivity sensors,environmental factors and hydrogen concentration are established respectively,and the error norm,relative error and coefficient of determination are introduced as the evaluation indicators of the model.The prediction performance of the model before and after the improvement of the BP algorithm is compared.The results show that the improved BP neural network has a better generalization ability for hydrogen concentration.Finally,this dissertation builds an experimental platform for hydrogen sensor detection system,and conducts a series of experimental analysis on the output characteristics,gas sensitivity repeatability,temperature and humidity characteristics and gas flow characteristics of the sensor.The prediction effects of hydrogen sensor detection system based on GA-BP neural network at 30? and 10? were tested respectively,and the average error of the prediction was only about 2%,which verified the reliability of the optimized detection system for high-precision identification of gas concentration under complex environment.