Advanced Soft Computing Techniques in Water Engineering: Comprehensive Insights into Applications, Trends, Cutting-Edge Algorithms

Abstract

Past literature in hydraulic structures has extensively used analytical and numerical models to analyze flow patterns, local scour, sediment transport, and energy dissipation in the system. in such cases, the presence of multi-phase and turbulent processes may result in highly complicated modeling processes that involve large computational requirements, whereas simpler approaches would compromise prediction accuracy. Nevertheless, artificial intelligence, machine learning, and soft computing tools have been extensively used for dealing with such issues in the recent past, but associated literature appears to be dispersed among various hydraulic sectors. Differing from the previous literature that addresses one isolated niche in hydraulics, the main scientific novelty in this study is the development of a comprehensive classification scheme together with diagnosis of existing research gaps. This will be achieved by means of using the systematic mapping approach, guided by the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines, which will involve the examination and synthesis of 281 primary studies published between 2000 and 2025. In terms of study selection, only original scientific journal articles indexed in Science Citation Index Expanded (SCIE) were considered for the period of 2000 to 2025. The proposed analytical model is based on an effective three-fold categorization criterion, which allows classifying the existing literature based on hydraulic systems, hydraulic processes, and algorithms in artificial intelligence. The analysis of frequency distribution and time dynamics within each category reveals important features that go far beyond the mere listing of existing approaches. Firstly, the taxonomical structure of artificial intelligence algorithms can be considered as a guide for the proper selection of a model depending on the complexity of the hydraulic processes. Secondly, the use of 176 different artificial intelligence algorithms across 48 hydraulic systems and 54 hydraulic phenomena can be considered as an important indicator of research saturation and undersaturation. For example, while channels and weirs are highly saturated, critical systems such as fishways or skimming walls remain severely under-investigated, comprising only one of the 281 primary studies reviewed. Among hydraulic systems, channel, weir, dam, river, pipe, and bridge are frequently studied, whereas skimming wall, bucket, rubble mound, district water system, fishway, and irrigation are the least studied. Among hydraulic phenomena, flow, source, sediment, discharge, scour, and hydraulic jump occupy the leading positions. Moreover, findings show that of the 176 techniques utilized, artificial neural network, multilayer perceptron, convolutional neural network, recurrent neural network, wavelet neural network, backpropagation, and radial basis function neural networks are highly adopted. Ultimately, this review clarifies methodological trends and highlights neglected domains, offering scholars and policymakers a clear framework to apply soft computing techniques to the most critically understudied areas in water engineering.