Wavelet and mixture of soft sensors to improve the monitoring of environmental parameters by neural network

Soft sensors, based on Elman NNs, have been developed to provide virtual measurements at different locations on the monument surface using as input source only the measurements acquired by an Air Ambient Monitor Station located nearby. Simulation of measurements by trained NN is a useful computational tool to monitor the physical or chemical conditions of the composing materials in a not invasive way, but their accuracy has to be high as analyzed from a metrological and statistical point of view. Two different mathematical and computational tools can be adopted to improve the accuracy of the virtual measurements: a wavelet preprocessing of times series data and the mixture soft sensors to fuse several input sources..