Augusto Montisci is associate professor in the scientific-disciplinary sector Electrotechnics. He received his PhD in mechanical design from the University of Cagliari (Italy). From 2002 he worked at the Department of Electrical and Electronic Engineering of the University of Cagliari as Researcher, and became Associate Professor in 2016. His teaching activity concerns circuit theory, signal processing and use of electromagnetic fields in the industrial field. He is a member of the teaching board of the Doctorate in Industrial Engineering at the University of Cagliari, where he contributes to teaching activities as a supervisor and giving courses for doctoral students. His scientific activity concerns Artificial Neural Networks, Optimization, Non-destructive tests, Fault diagnosis, Renewable energy sources, Magnetohydrodynamics. He is the author of about 82 indexed works, which received more than 380 citations, obtaining an h-index of 11 (Scopus database). His publishing activity includes the review of numerous international magazines and conferences, he was a member of the technical committee of international conferences, where he was chairman of sessions. He was chairman of the International Conference on Fundamentals and Applied MHD (PAMIR 2016). He was a member of the founding committee of the European Society of Magnetohydrodynamics (EuMHD) which he currently holds the position of president.

He has participated in over 20 international and national projects, in many of which he has played a coordinating role:

1. FSE della Comunità Europea, Regione Autonoma della Sardegna: “SINEQUAL sviluppo di un SIstema NEurale per il controllo di QUALità di circuiti stampati”,

2. Progetto di cooperazione internazionale finanziato dalla L.R. n. 19/96, partner University of Oran, Algeria, 2001: “Controllo di qualità dei prodotti dell’industria agro-alimentare in Algeria”.

3. Programma di ricerca in collaborazione con i paesi in via di sviluppo, finanziato dalla L.R. 43/90 e L.R. 26/96, partner University of NoviSad, Yugoslavia, 2001: “Decision Trees in Machine learning”.

5. Progetto di cooperazione internazionale finanziato dalla L.R. n. 19/96, partner University of Oran, Algeria, 2002: “Controllo di qualità dei prodotti dell’industria agro-alimentare in Algeria”.

6. Collaborazione Italo-Britannica per la ricerca e l’istruzione superiore, finanziato da CRUI e British Council, partner University of Leeds, UK. 2002-03: “Forecasting Hydrological Series using Neural Networks”

7. Progetto di cooperazione internazionale finanziato dalla L.R. n. 19/96, partner University of Oran, Algeria, partner University of Science and Technology of Oran (Algeria), e Istituto per il Monitoraggio degli Agrosistemi – CNR di Sassari, 2003-2005: “Diffusione e trasferimento di tecnologie “smart” per la gestione delle risorse idriche ed alimentari in Algeria”

8. PRIN 2003: “Studio e sperimentazione di diagnostiche non distruttive per tratti non accessibili di condutture”

9. POR Sardegna 2000-6, Mis. 3.13: “Sistemi Automatici Multisensoriali per l’Analisi del Comportamento dell’Animale da laboratorio”

10. PRIN 2007 “Studio e sperimentazione di un sistema di comunicazione ad onde convogliate in impianti elettrici navali”

11. PRIN 2009 “Diagnostica non distruttiva ad ultrasuoni tramite sequenze pseudo-ortogonali per imaging e classificazione automatica di prodotti industriali”

12. ENPI CBCMED 2012 “LANDCARE MEDiterranean cross-border network for local rural governance improvement to enhance rural waste management”

13. POR FSER 2007-13 “Realizzazione di un sistema innovativo di acquisizione e trascrizione automatica di partiture musicali da strumenti acustici tradizionali ed etnici con una particolare attenzione verso le Launeddas”

14. EUROPEAID 2017 "Development of a Model Municipal Solid Waste Management Program for the Protection of the Saniq River Basin in Southem Lebanon"

The scientific activity carried out mainly concerns the following fields:

Diagnosis using Neural Networks Artificial neural networks are used to solve fault diagnosis and defect problems, formulating the diagnosis in terms of classification problem. The methodology was applied to different types of systems, such as electrical circuits, for which deviation faults are diagnosed and catastrophic faults, printed circuits, for which the diagnosis is made for visual inspection of assembly defects, underground pipes, where the faults consist of erosion points and are detected by recognizing the sound waves reflected at an inspection point.

Synthesis Algorithms of Artificial Neural Networks The possibility of effectively using artificial neural networks in problems of classification, regression, prediction, parametric identification, design, inversion, depends critically on the degree of reliability of the neural network on which the specific method is based. For this reason we proceeded to the development of algorithms for the synthesis of neural networks that would allow us to establish the degree of precision of the neural network produced in a deterministic way. We have first developed a method that allows us to develop neural classifiers without errors on the training set, after which we have extended the method for the construction of neural networks with non-linear regression functions, for which it is possible to establish a priori the precision at the training examples. These algorithms are based on linear programming techniques and for this reason they are exempt from trapping problems in local minima, typical of gradient-based learning algorithms. Finally, a training technique for locally recurrent neural networks was developed in which an algorithm based on the error gradient is used, explicitly calculated with respect to all the parameters of the network, including the feedback parameters.

Fault diagnosis based on Testability analysis In addition to the approach to diagnosis defined as a point classification problem, in the particular case of the diagnosis of analog circuits, a diagnostic method has been developed that takes advantage of the a priori Testability analysis of circuit, to define coherent training sets for neural networks. With this setting, neural networks become a means to obtain the solution of fault equations without having to solve them analytically. The testability analysis is essentially used to define ambiguity groups, i.e. those fault configurations that cannot be distinguished mutually only on the basis of measurements. For this purpose, methods found in the literature have been used, based on the linearization of the fault equation system, but an alternative method has also been proposed, in which the condition that the extent of the fault is such as to allow a linearization of the problem is not assumed. This approach uses Gröbner's bases and can be applied in symbolic terms, therefore with a modest computational cost.

Optimization and inversion methods based on neural networks Optimization techniques based on neural networks have been developed. In these techniques the calculation of the objective function is carried out for a limited set of examples using finite element techniques, therefore with a very high computational burden. Exploiting this set, a neural model is built, which allows to replace the finite elements, making the inspection of the objective function much faster. The optimization method exploits the algebraic structure of the neural network that realizes the approximation of the system under examination. In fact, it consists in solving an inverse problem, in which the sought value of the objective function is imposed on a neural network, and the corresponding input value is determined, which represents the solution to the optimization problem. The precision of the neural model that calculates the objective function is crucial for the goodness of the solution that can be obtained. To this end, it has benefited from the synthesis techniques of the neural networks described above.

Non-Destructive Testing Non-destructive techniques (NDT) have been applied to the diagnosis of metal equipment and valuable masonry works. As a test signal, ultrasounds were used, applied and detected by piezometric transducers, which combine the excellent operating parameters with the possibility of easily controlling the characteristics of the excitation signal. The developed technique involves the realization of a reference set for the training of a neural model of the system, in order to greatly limit the burden of the test tests. The system under test, whether it be a metallic device or a masonry, is simulated by means of a finite element code so as to be able to determine the behavior in the presence of different types of defect, which can somehow be considered canonical. A certain number of field investigations are necessary to carry out the calibration of the model and its subsequent validation. On the basis of a significant number of examples in simulation it is possible to define a sequence of operations that allows to carry out tests on the real object. Several techniques have been compared for the resolution of the diagnosis problem, among them artificial neural networks have been used, methods based on the study of the characteristics of the acquired signal, and finally tomographic inversion techniques, for which algorithms have been developed original.

Powerline data transmission The transmission of signals on power lines has been a technique in use for several decades, but until a few years ago the use was limited to the transmission of control signals. Recently the technique has been extended to the transmission of audio and video signals, which has posed a series of problems related to interference, transmission speed, bandwidth, and many others. To identify viable technical solutions to tackle these problems, approaches have been developed that exploit the skills acquired in other areas and which are described in the previous points. In particular, multi-objective optimization techniques were used to identify the best compromise solutions between the conflicting needs to increase the transmission rate (Bit Rate), to limit the error rate (Bit Error Rate) and to limit the power used. Subsequently the problem of the limitation of power transmission peaks (Peak-to-Average Power Ratio) was faced, which represent a serious problem in the management of power line systems both because they cause interference on the surrounding equipment, and because they can compromise the integrity of the transmitted signal.

Innovative Systems for Electricity Production In order to increase the energy conversion efficiency or to make possible an increase in the exploitation of renewable sources, innovative systems have been proposed for the production of electricity. In particular, an induction magnetohydrodynamic electric generator (MHD) and an integrated system in the building for the exploitation of wind energy in urban areas were presented. The first system represents an evolution with respect to the functional scheme of the MHD generators that aroused great interest during the 1960s, but which were then progressively abandoned in favor of other conversion systems, such as the gas turbines that had undergone a strong impulse from part of the aviation industry. The functional induction scheme arises as an alternative to that of the old MHD generators and in fact eliminates the limits that have affected its use on a large scale, i.e. the need to work at very high temperatures, to generate very intense magnetic fields, to add the operating fluid with additives that increase its conductivity.

The system called "wind roof" at the same time performs the function of a roof like a normal roof and acts as a flow conveyor that converges the wind towards a centripetal turbine with a vertical axis placed at the center of the system. The particular conformation of the conveying ducts is designed in such a way as to limit air turbulence and moreover the flow takes advantage of the depression areas that are created above the building and in the wall opposite to that exposed to the wind. The vertical axis turbine makes it possible to do without complex blades orientation systems, and the reduced dimensions compared to the intercepted air flow cross section, allow to obtain numerous advantages, such as a low specific installation and maintenance cost, a wide range of wind speeds compatible with turbine operation, low noise levels. Finally, the fact that the generation system is invisible to the outside avoids any urban planning problem. These features offer interesting perspectives, both from a technical and an economic point of view, for the self-production of electricity from renewable sources in residential areas.

 

Principali Pubblicazioni

1.        FODDIS M.L., MONTISCI A., TRABELSI F., URAS G., An MLP-ANN-based approach for assessing nitrate contamination (2019) WATER SCIENCE AND TECHNOLOGY: WATER SUPPLY, Vol. 19 (7)

2.        DELOGU R.S., MONTISCI A., PIMAZZONI A., SERIANNI G., SIAS G., Neural network based prediction of heat flux profiles on STRIKE (2019) FUSION ENGINEERING AND DESIGN, Vol. 146

3.        CARCANGIU S., FANNI A., MONTISCI A., Electric capacitance tomography for nondestructive testing of standing trees (2019) INTERNATIONAL JOURNAL OF NUMERICAL MODELLING: ELECTRONIC NETWORKS, DEVICES AND FIELDS, Vol. 32 (4)

4.        CARCANGIU S., FANNI A., MONTISCI A., Optimization of a power line communication system to manage electric vehicle charging stations in a smart grid (2019) ENERGIES, Vol. 12 (9)

5.        CANNAS B., CARCANGIU S., FANNI A., FARLEY T., MILITELLO F., MONTISCI A., PISANO F., SIAS G., WALKDEN N., Towards an automatic filament detector with a Faster R-CNN on MAST-U (2019) FUSION ENGINEERING AND DESIGN

6.        ALSARI M., PEARSON A.J., WANG J.T.-W., WANG Z., MONTISCI A., GREENHAM N.C., SNAITH H.J., LILLIU S., FRIEND R.H., Degradation Kinetics of Inverted Perovskite Solar Cells (2018) SCIENTIFIC REPORTS, Vol. 8 (1)

7.        CARCANGIU S., FANNI A., FORCINETTI R., MONTISCI A., Multiobjective Tabu Search algorithm for the optimal design of a Thermo-Acoustic Magneto-Hydro-Dynamic electric generator (2018) INTERNATIONAL JOURNAL OF APPLIED ELECTROMAGNETICS AND MECHANICS, Vol. 56 (S1)

8.        ROUX J.-P., ALEMANY A., MONTISCI A., Thermoacoustic Magnetohydrodynamic Electric Generator (2017), EU pat. EP2874292 (B1), Bulletin 2017/39

9.        CARCANGIU S., FORCINETTI R., MONTISCI A., Simulink model of an iductive MHD generator (2017) MAGNETOHYDRODYNAMICS, Vol. 53 (2)

10.      ALEMANY A., FORCINETTI R., MASSON F., MONTISCI A., FEM analysis of the inflation process of magnetoplasma sails (2017) MAGNETOHYDRODYNAMICS, Vol. 53 (2)

11.      FODDIS M.L., MATZEU A., MONTISCI A., URAS G., The Arborea plain (Sardinia - Italy) nitrate pollution evaluation (2017) ITALIAN JOURNAL OF ENGINEERING GEOLOGY AND ENVIRONMENT, Vol. 2017 (Specialissue1)

12.      FODDIS M.L., ACKERER P., MONTISCI A., URAS G., ANN-based approach for the estimation of aquifer pollutant source behaviour (2015) WATER SCIENCE AND TECHNOLOGY: WATER SUPPLY, Vol. 15 (6)

13.      FODDIS M.L., MATZEU A., MONTISCI A., URAS G., Application of three different methods to evaluate the nitrate pollution of groundwater in the Arborea plain (Sardinia - Italy) (2015) RENDICONTI ONLINE SOCIETA GEOLOGICA ITALIANA, Vol. 35

14.      ALEMANY A., CARCANGIU S., FORCINETTI R., MONTISCI A., ROUX J.P., Feasibility analysis of an MHD inductive generator coupled with a thermoacoustic resonator (2015) MAGNETOHYDRODYNAMICS, Vol. 51 (3)

15.      CARCANGIU S., MONTISCI A., PINTUS R., Performance analysis of an inductive MHD generator (2012) MAGNETOHYDRODYNAMICS, Vol. 48 (1)

16.      CARCANGIU S., MONTISCI A., Performance assessment of an aeolian roof for the exploitation of wind power in urban areas (2012) RENEWABLE ENERGY AND POWER QUALITY JOURNAL, Vol. 1 (10)

17.      CANNAS B., DELOGU R.S., FANNI A., MONTISCI A., SONATO P., ZEDDA K., Geometrical kernel machine for prediction and novelty detection of disruptive events in TOKAMAK machines (2010) JOURNAL OF SIGNAL PROCESSING SYSTEMS, Vol. 61 (1)

18.      CARCANGIU S., FANNI A., MEREU A., MONTISCI A., Grid-enabled tabu search for electromagnetic optimization problems (2010) IEEE TRANSACTIONS ON MAGNETICS, Vol. 46 (8)

19.      CANNAS B., FANNI A., MONTISCI A., Algebraic approach to ambiguity-group determination in nonlinear analog circuits (2010) IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I: REGULAR PAPERS, Vol. 57 (2)

20.      CARCANGIU S., FANNI A., MONTISCI A., A constructive algorithm of neural approximation models for optimization problems (2009) COMPEL - THE INTERNATIONAL JOURNAL FOR COMPUTATION AND MATHEMATICS IN ELECTRICAL AND ELECTRONIC ENGINEERING, Vol. 28 (5)

21.      CARCANGIU S., FANNI A., MONTISCI A., Multiobjective tabu search algorithms for optimal design of electromagnetic devices (2008) IEEE TRANSACTIONS ON MAGNETICS, Vol. 44 (6)

22.      DELOGU R., FANNI A., MONTISCI A., Geometrical synthesis of MLP neural networks (2008) NEUROCOMPUTING, Vol. 71 (4-6)

23.      CARCANGIU S., DI BARBA P., FANNI A., MOGNASCHI M.E., MONTISCI A., Comparison of multi-objective optimisation approaches for inverse magnetostatic problems (2007) COMPEL - THE INTERNATIONAL JOURNAL FOR COMPUTATION AND MATHEMATICS IN ELECTRICAL AND ELECTRONIC ENGINEERING, Vol. 26 (2)

24.      CAU F., DI MAURO M., FANNI A., MONTISCI A., TESTONI P., A neural networks inversion-based algorithm for multiobjective design of a high-field superconducting dipole magnet (2007) IEEE TRANSACTIONS ON MAGNETICS, Vol. 43 (4)

25.      CAU F., FANNI A., MONTISCI A., TESTONI P., USAI M., A signal-processing tool for non-destructive testing of inaccessible pipes (2006) ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE, Vol. 19 (7)

26.      CANNAS B., CAU F., FANNI A., MONTISCI A., TESTONI P., USAI M., Neural NDT by means of reflected longitudinal and torsional waves modes in long and inaccessible pipes (2005) WSEAS TRANSACTIONS ON SYSTEMS, Vol. 4 (11)

27.      CHERUBINI D., FANNI A., MONTISCI A., TESTONI P., A fast algorithm for inversion of MLP networks in design problems (2005) COMPEL - THE INTERNATIONAL JOURNAL FOR COMPUTATION AND MATHEMATICS IN ELECTRICAL AND ELECTRONIC ENGINEERING, Vol. 24 (3)

28.      CHERUBIM D., FANNI A., MONTISCI A., TESTONI P., Inversion of MLP neural networks for direct solution of inverse problems (2005) IEEE TRANSACTIONS ON MAGNETICS, Vol. 41 (5)

29.      FANNI A., MONTISCI A., A neural inverse problem approach for optimal design (2003) IEEE TRANSACTIONS ON MAGNETICS, Vol. 39 (3 I)

30.      FANNI A., GIUA A., MARCHESI M., MONTISCI A., Neural network diagnosis approach for analog circuits (1999) APPLIED INTELLIGENCE, Vol. 11 (2)