Development of Chebyshev neural network-based smart sensors for noisy harsh environment

Smart sensing of environmental parameters is an important task in robotics, process industries, sensor networks and autonomous systems. In this paper, we propose a novel Chebyshev neural network (ChNN) to develop smart sensors which can provide linearized and accurate readout, and can compensate for nonlinear environmental disturbances including additive noise. By taking two environmental models and using a capacitive pressure sensor as an example, through computer simulations, performance comparison was carried out between the proposed ChNN and Multilayer Perceptron (MLP)-based sensor models over a wide temperature range and additive noise. We have shown that the performances in terms of full scale error and sensor readout of both the NNs are similar. But a major advantage of the ChNN is that due to its single-layer architecture it provides substantial computational advantage over MLP.