Here is reported about the design of a Soft Sensor (SS) able to monitor hazardous gases in industrial plants. The SS is designed to estimate the gas concentrations by means of the measurements coming from an array of sensors, avoiding at the same time the humidity and temperature influence on array outputs. The SS has been designed with a data-driven approach, using a set of experimental data acquired in lab. Two different SSs design methodologies are compared, with the aim of obtaining both good performance and a low computational complexity. As a first approach, a Principal Component Analysis (PCA) has been performed to exploit the high correlation among some of the measures coming from the sensor array. A classical Multi-Layer Perceptron neural network is then trained to estimate the relationships between the PCA outputs and the gas concentrations. As a second approach, a Deep Belief Network (DBN), has been considered. The data here reported show good accuracy in the evaluation of several gas concentrations, even in presence of noised measurements, allowing an efficient risk warning. Even if both methods gave similar performance, a lower number of model parameters and a lower noise sensitivity is obtained when using DBNs.

This paper, for the first time, analyzes in detail the kink phenomenon in S₂₂ as observed in GaN HEMT technology. To gain a comprehensive understanding, the kink effect (KE) is studied with respect to temperature and bias conditions. The achieved results clearly show that the dependence of the KE on the operating condition should be mainly ascribed to the transconductance, which plays a determinant role in the appearance of this effect. Furthermore, the analysis is extended to investigate the peak in the magnitude of h₂₁ showing its disappearance at low drain-source voltage, due to the increase of the intrinsic output conductance. The importance of this investigation originates from the fact that an accurate and complete characterization of these anomalous phenomena enables microwave engineers to properly take them into account during the modeling and design phases.enables microwave engineers to properly take them into account during the modeling and design phases.

Neural networks are proposed for efficient temperature-dependent modeling of small-signal and noise performances of low-noise microwave transistors over a wide temperature range. The proposed models can be based either on neural networks only or on a combination of neural networks and empirical transistor models. © 2005 Wiley Periodicals, Inc. Int J RF and Microwave CAE, 2005.