Meteorological observation plays an important role in establishing atmospheric theory and improving the accuracy of weather forecasts. Sensors are the foundation and the core in the meteorological observation. Only the accurate calibration of the sensor can ensure the validity of measurement data. In most current methods of sensor calibration, the least squares method is used for calibration which results in low calibration accuracy. In this paper, a sensor calibration model is implemented by using artificial intelligence technology. By combining a back propagation (BP) neural network, a Gaussian function and the Levenberg-Marquardt (LM) algorithm, an enhanced BP network is realized for sensor calibration. The calibration model is transplanted to a Microcontroller unit (MCU). The Gaussian function is fitted by piecewise polynomials, which effectively reduces the computing resources and time use of the MCU. The experimental results show that: the traditional BP network reduces the mean squared error (MSE) of the atmospheric pressure sensor from 5.93 to 2.83. This reduces the measurement error by 52.28%. The enhanced BP network reduces the MSE to 0.77, which further reduces the measurement error by 34.74%. By fitting a Gaussian function to a polynomial, the computing time of MCU is reduced from 40 us to 0.1 us, which significantly reduces the complexity of the algorithm.

No abstract

Temperature effects on current transport and the negative differential resistance for SiGe/Si and GaAlAs/GaAs resonant tunneling structures (RTS) have been studied, and the maximum working temperature Tm has been estimated. The calculations show that decreases in the carrier effective mass, well width and barrier thicknesses, lead to better temperature characteristics, implying higher peak current and larger peak-to-valley ratio (PVR) at higher temperature. These results are consistent with our experiment on SiGe/Si RTSs and other published experiments. The crucial role of nonresonant tunneling current Jnon in temperature effects for current transport is emphasized. Suggestions for optimizing RTS design to increase its Tm and PVR are discussed.