The potential of two Kohonen artificial neural networks I ANNs) - linear vector quantisa - tion (LVQ) and the self organising map (SOM) - is explored for pulse shape discrimination (PSD), i.e. for distinguishing between neutrons (n's) and gamma rays (γ’s). The effect that la) the energy level, and lb) the relative- of the training and lest sets, have on iden- tification accuracy is also evaluated on the given PSD datasel The two Kohonen ANNs demonstrate compfcmentary discrimination ability on the training and test sets: while the LVQ is consistently mote accurate on classifying the training set. the SOM exhibits higher n/γ identification rales when classifying new paltms regardless of the proportion of training and test set patterns at the different energy levels: the average tint: for decision making equals ∼100 /e in the cax of the LVQ and ∼450 μs in the case of the SOM.

A computational intelligence-based identification of the properties of maximally sustainable materials for a given application, as derived from key properties of existing candidate materials, is put forward. The correlation surface between material properties (input) and environmental impact (EI) values (output) of the candidate materials is initially created using general regression (GR) artificial neural networks (ANNs). Genetic algorithms (GAs) are subsequently employed for swiftly identifying the minimum point of the correlation surface, thus exposing the properties of the maximally sustainable material. The ANN is compared to and found to be more accurate than classic polynomial regression (PR) interpolation/prediction, with sensitivity and multicriteria analyses further confirming the stability of the proposed methodology under variations in the properties of the materials as well as the relative importance values assigned to the input properties. A nominal demonstration concerning material selection for manufacturing maximally sustainable liquid containers is presented, showing that by appropriately picking the pertinent input properties and the desired material selection criteria, the proposed methodology can be applied to a wide range of material selection tasks.

No abstract

An artificial neural network (ANN) is employed for sorting a sequence of real elements in monotonic (descending or ascending) order. Although inspired by harmony theory (HT), whereby the same construction as for the HT ANN is followed, the proposed ANN differs in the mode of operation, namely the obliteration of the consensus (harmony) function, the circumvention of simulated annealing as a means of settling to a solution, the simplification of the activation updating of the nodes of the upper layer, the clamping of the nodes of the lower layer, the gradual shrinking of the ANN and the use of an automatic termination criterion. The creation of the sorted sequence is progressive, whereby at most as many network updates are required as there are elements in the sequence. Ties between elements are resolved by simultaneous activation of the corresponding nodes. Finally, the min and max problems are solved in a single network update.

In the guise of artificial neural networks (ANNs), genetic/evolutionary computation algorithms (GAs/ECAs), fuzzy logic (FL) inference systems (FLIS) and their variants as well as combinations, the computational intelligence (CI) paradigm has been applied to nuclear energy (NE) since the late 1980s as a set of efficient and accurate, non-parametric, robust-to-noise as well as to-missing-information, non-invasive on-line tools for monitoring, predicting and overall controlling nuclear (power) plant (N(P)P) operation. Since then, the resulting CI-based implementations have afforded increasingly reliable as well as robust performance, demonstrating their potential as either stand-alone tools, or - whenever more advantageous - combined with each other as well as with traditional signal processing techniques. The present review is focused upon the application of CI methodologies to the - generally acknowledged as - key-issues of N(P)P operation, namely: control, diagnostics and fault detection, monitoring, N(P)P operations, proliferation and resistance applications, sensor and component reliability, spectroscopy, fusion supporting operations, as these have been reported in the relevant primary literature for the period 1990-2015. At one end, 1990 constitutes the beginning of the actual implementation of innovative, and – at the same time – robust as well as practical, directly implementable in H/W, CI-based solutions/tools which have proved to be significantly superior to the traditional as well as the artificial-intelligence-(AI)derived methodologies in terms of operation efficiency as well as robustness-to-noise and/or otherwise distorted/missing information. At the other end, 2015 marks a paradigm shift in terms of the emergent (and, swiftly, ubiquitous) use of deep neural networks (DNNs) over existing ANN architectures and FL problem representations, thus dovetailing the increasing requirements of the era of complex - as well as Big - Data and forever changing the means of ANN/neuro-fuzzy construction and application/performance. By exposing the prevalent CI-based tools for each key-issue of N(P)P operation, overall as well as over time for the given 1990-2015 period, the applicability and optimal use of CI tools to NE problems is revealed, thus providing the necessary know-how concerning crucial decisions that need to be made for the increasingly efficient as well as safe exploitation of NE.