L. Bertucco

A rigorous inter-comparison of ground-level ozone predictions

Novel statistical approaches to prediction have recently been shown to perform well in several scientific fields but have not, until now, been comprehensively evaluated for predicting air pollution. In this paper we report on a model inter-comparison exercise in which 15 different statistical techniques for ozone forecasting were applied to ten data sets representing different meteorological and emission conditions throughout Europe. We also attempt to compare the performance of the statistical techniques with a deterministic chemical trajectory model. Likewise, our exercise includes comparisons of sites, performance indices, forecasting horizons, etc. The comparative evaluation of forecasting performance (benchmarking) produced 1340 yearly time series of daily predictions and the results are described in terms of predefined performance indices. Through analysing associations between the performance indices, we found that the success index is of outstanding significance. For models that are excellent in predicting threshold exceedances and have a high success index, we also observe high performance in the overall goodness of fit. The 8-h average ozone concentration forecast accuracy was found to be superior to the 1-h mean ozone concentration forecast, which makes the former very significant for operational forecasting. The best forecasts were achieved for sites located in rural and suburban areas in Central Europe unaffected by extreme emissions (e.g. from industries). Our results demonstrate that a particular technique is often excellent in some respects but poor in others. For most situations, we recommend neural network and generalised additive models as the best compromise, as these can handle nonlinear associations and can be easily adapted to site specific conditions. In contrast, nonlinear modelling of the dynamical development of univariate ozone time-series was not profitable.

A cellular neural networks approach to flame image analysis for combustion monitoring

Proposes an approach based on cellular neural networks (CNNs) to the analysis of flame images for real time monitoring of combustion process in a waste incinerator. The use of CNNs analysis is dictated by the high images sampling rate, which was necessary due to the fast dynamics of the process under study. The dynamical behavior of the descriptors of the images processed by the CNNs was also studied and the results of this analysis are also presented.

A cellular neural network based system for cell counting in culture of biological cells

Cell counting methods are important tools in molecular biology as well as clinical medicine. It is not always technically possible to measure quantitatively the events of cellular growth and fission. When it can be done, the procedures are neither so simple nor without excessive tedium as to lend themselves practically to the necessary replication of observations with large number of individual cells. We describe a CNN based system that uses a CNN simulator for counting cells. The performances of the proposed system are illustrated by a simple cell counting experiment using a Petroff-Hauser based counter system.

A neural approach to the integrated inversion of geophysical data of different types

Artificial neural networks (ANNs) have been employed for the inversion of the geometrical parameters of a magma-filled dike, which causes observable changes in various geophysical fields. The inversion approach, which is based on the function approximation capabilities of multilayer perceptrons (MLPs), is also carried out by a systematic search technique based on the simulated annealing (SA) optimization algorithm in order to emphasize the merits of the proposed strategy. It is shown that even if the SA approach guarantees a high degree of accuracy, it requires a considerable amount of time, incompatible with on-line applications. On the other hand, it is shown that MLPs, once correctly trained, can solve the inversion problem very fast and with an appreciable degree of accuracy. It is also demonstrated that an integrated approach involving geophysical data of different kinds allows for a more accurate solution than when ground deformation data alone is considered. The results given in the paper are supported by experiments carried out using an interactive software tool developed ad hoc, which allows both direct and inverse modeling of data related to the opening of a crack at the beginning and throughout a volcanic activity episode.

CNN with non-integer order cells

A new kind of cellular neural network (CNN) is introduced. Its feature consists of a state representation using q-order derivatives, with q being a non-integer quantity. This approach can be considered as a generalisation of the traditional CNN model, which is obtained from the one presented in the paper as a particular case setting q=1. It is shown that this more general CNN structure exhibits suitable performance in terms of processing speed. Various examples are reported to show the suitability of non-integer order CNNs.

Predicting Daily Average SO2 Concentrations in the Industrial Area of Syracuse (Italy)

In this paper artificial neural networks are used to build 1- day-ahead SO2 prediction models. The structure of the model was obtained following appropriate statistical analysis of the time series.

Autowaves in noninteger order CNNs

In this paper it is shown that complex spatio-temporal phenomena, usually met in physical and biological systems, can be reproduced by means of cellular neural networks of noninteger order. The template parameters are reported in the paper, together with some simulation results which show the suitability of the approach.

Soft computing techniques for modeling geophysical data

The inversion problem dealt with is identification of the parameters of a magma-filled dike which causes observable changes in various geophysical fields, using artificial neural networks (ANNs). The inversion approach, which is based on the function approximation capabilities of multi-layer perceptrons (MLPs), is also carried out by a systematic search technique based on the simulated annealing (SA) optimization algorithm, in order to emphasize the peculiarities of the proposed strategy. In the paper it is demonstrated that MLPs, once correctly trained, can solve the inversion problem very fast with an appreciable degree of accuracy. It also demonstrated that an integrated approach involving geophysical data of different types, allows for a more accurate solution than when only ground deformation data is considered.

A cellular neural networks approach for non-destructive control of mechanical parts

An approach is proposed using cellular neural networks applied image processing, for the detection and characterisation of superficial faults in mechanical parts. There are above all two advantages deriving from an application of the proposed methodologies: the automization of a procedure, that of non-destructive tests (NDT), which is today carried out manually, and the possibility to reduce to a negligible amount the time spent on checking operations at present estimated to be in the order of a number of hours for each separate mechanical part.