Machine learning (ML) methods have shown powerful performance in different application Nonetheless, designing ML models remains a challenge and requires further research as most procedures adopt a trial and error strategy In this study, we present a methodology to optimize the architecture and the feature configurations of ML models considering a supervised learning process The proposed approach employs genetic algorithm (GA)-based integer-valued optimization for two ML models, namely deep neural networks (DNN) and adaptive neuro-fuzzy inference system (ANFIS) The selected variables in the DNN optimization problems are the number of hidden layers, their number of neurons and their activation function, while the type and the number of membership functions are the design variables in the ANFIS optimization problem The mean squared error (MSE) between the predictions and the target outputs is minimized as the optimization fitness function The proposed scheme is validated through a case study of computational material design We apply the method to predict the fracture energy of polymer/nanoparticles composites (PNCs) with a database gathered from the literature The optimized DNN model shows superior prediction accuracy compared to the classical one-hidden layer network Also, it outperforms ANFIS with significantly lower number of generations in GA The proposed method can be easily extended to optimize similar architecture properties of ML models in various complex systems

/pdf/an-efficient-optimization-approach-for-designing-machine-3u2y0aqzlx.pdf

An efficient optimization approach for designing machine learning models based on genetic algorithm

The mechanical behaviour of drainage pipeline under traffic load before and after polymer grouting trenchless repairing

To effectively stabilize coarse sand, bentonite was introduced in microbially induced carbonate precipitation (MICP) grouting. Varying concentrations (0, 20, 40, and 80 g/L) of bentonite were added...

Strength and permeability of bentonite-assisted biocemented coarse sand

Due to the rapid growth of 3D printing popularity, including fused deposition modeling (FDM), as one of the most common technologies, the proper understanding of the process and influence of its parameters on resulting products is crucial for its development. One of the most crucial parameters of FDM printing is the raster angle and mutual arrangement of the following filament layers. Presented research work aims to evaluate different raster angles (45°, 55°, 55'°, 60° and 90°) on the static, as well as rarely investigated, dynamic mechanical properties of 3D printed acrylonitrile butadiene styrene (ABS) materials. Configuration named 55'° was based on the optimal winding angle in filament-wound pipes, which provides them exceptional mechanical performance and durability. Also in the case of 3D printed samples, it resulted in the best impact strength, comparing to other raster angles, despite relatively weaker tensile performance. Interestingly, all 3D printed samples showed surprisingly high values of impact strength considering their calculated brittleness, which provides new insights into understanding the mechanical performance of 3D printed structures. Simultaneously, it proves that, despite extensive research works related to FDM technology, there is still a lot of investigation required for a proper understanding of this process.

Fuming Wang

Papers

The mechanical behaviour of drainage pipeline under traffic load before and after polymer grouting trenchless repairing

Study on the mechanical properties of polyurethane (PU) grouting material of different geometric sizes under uniaxial compression

Numerical simulation and full-scale test on dynamic response of corroded concrete pipelines under Multi-field coupling

Experimental and practical investigation of the sealing efficiency of cement grouting in tortuous fractures with flowing water

Experimental and numerical study on polymer grouting pretreatment technology in void and corroded concrete pipes