Study On The Dynamic Performance Of The Sensor

The sensor is the nerve endings of the Internet of Things,it is also the core components for human beings to perceive the whole world.To improve the measurement accuracy of the sensor,we need to study the performance of the sensor in-depth, and the dynamic performance of the sensor is very important to its measurement accuracy. So this paper conducted a series of theoretical and experimental work related to the dynamic performance of the sensor including the modeling of the dynamic nonlinear of the sensor, the analysis of the sensor dynamic performance and the calibration of the dynamic nonlinear performance.The study of the sensor dynamic performance is based on accurate sensor dynamic models, and aimed at the improvement of its dynamic performance. In order to improve the modeling and compensating precision, this paper proposed a novel variable step-size&variable parameters least mean square error adaptive algorithm who can be extended to nonlinear. This algorithm can be used in signal filtering and system identification in the study of the sensor dynamic performance. More over, a modified decorrelation principle was also proposed to figure out the problem of the system identification with correlated input signals. The analytic and simulation results verified the superiorities of our algorithm.To improve the sensor performance and taking use of the merits of the block oriented models, we chose the Wiener model to represent the model of the non-linear sensor. Based on the adaptive algorithm above and the idea of the neural network, a new two-step dynamic nonlinear adaptive neural network system identification algorithm based on the Wiener model was proposed, whose identify and training process was given. The simulation of the sensor by the combination of Lab VIEW and Matlab verified the correctness of the description of the sensor model and its identification algorithm.After correctly analyze the performance standards of the sensor, a two-step dynamic nonlinear adaptive neural network compensation method based on the Wiener model was proposed. System identification algorithm was used to compensate the two parts of the Wiener model. To solve the problem of dynamic compensation of the actual system may lead to high-frequency noise, the article also raised the idea of adaptive decision-making. Simulation results verified the correctness of our compensation and adaptive decision-making ideas. In real applications, we can decide whether to adopt the adaptive decision-making idea according to the performance of the system to be compensated.According to the working principle of the hot-film Mass Air Flow sensor, static and dynamic calibration experiments of it were designed. Pre-processing the experimental data and then applying the system modeling and compensation methods that have been verified effectively into the study of its dynamic performance of the sensor can correctly describe the dynamic nonlinear model of the sensor and improve its dynamic performance.