Design Of Fully Automatic Calibration System For Automotive Pressure Senso

With the development of science and technology,MEMS pressure sensors have been everywhere in life,among which automotive electronics is the largest application market of MEMS pressure sensors.When using it,the temperature change will affect its output accuracy.Compared with pressure sensors in other fields,automotive pressure sensor requires higher precision,therefore a lot of temperature compensation methods of pressure sensors are proposed.In this paper,taking the accuracy requirements of automotive pressure sensors as a standard,a pressure sensor automatic calibration system is designed based on improving GABP neural network.Firstly,aiming at the temperature compensation for the automotive pressure sensor,the improved self-adaptive genetic algorithm is forwarded.It applies Tent chaotic map to initialize the population,which greatly reduces the appearance of individuals who exceed the optimal solution.Taking half of the maximum fitness as the evaluation criterion of genetic operators avoids the problem that traditional adaptive genetic algorithms fall into local optimality due to the increase of the maximum fitness individuals.T-distribution is used to create a new generation of population by disturbing all individuals.The algorithm is used to optimize the BP neural network in order to establish the temperature compensation model of the improved GABP neural network.Through simulation experiment,it proves that the convergence speed and output accuracy of the model are both better than that of the BP neural network model.Secondly,aiming at the problem of waste of manpower and material resources in batch calibration of pressure sensors,a set of automatic calibration upper computer based on LabVIEW is designed.The upper computer mainly realizes the automatic control of the experimental platform,the collection and transmission of the original data of the pressure sensor,as well as the training of the neural network model and the preservation of the calibration parameters.At the same time,a communication acquisition board circuit is designed to realize the information interaction between the pressure sensor module and the upper computer,which can carry 8 pressure sensor modules simultaneously.Circuits can also be connected to each other to achieve batch calibration.In order to run the activation function of neural network in the conditioning chip,a table-driven linear fitting method with non-average segmentation is introduced.Simulation experiments and hardware tests prove that the activation function fitted by this method has high accuracy and could be run in the conditioning chip.Finally,an experimental platform is built in the laboratory and the designed calibration system is used to calibrate the pressure sensor module.The experimental data show that after calibration,the temperature drift of the pressure sensor reduces by an order of magnitude,and the output accuracy increases from 7.1%FS to 0.2%FS,which fully meets the design accuracy requirements of the automotive pressure sensor.