Neural network-aided prediction of post-cracking tensile strength of fibre-reinforced concrete

The need for reliable test data to verify scientific theories together with the necessity to calibrate code‐oriented expressions and their level of safety have pushed researchers to develop test databases. Such databases collect the work of multiple independent scientific initiatives compiling experimental evidences on similar test configurations but with different parameters such as geometry, size of the component tested, boundary conditions, and/or mechanical properties of the materials. Such databases, usually collected for specific purposes, have traditionally faced maintenance, and especially consistency issues. Also, they have given rise to conflicting values due to different interpretations of the same data entries in databases elaborated by different researchers. In addition, such databases are frequently designed and configured to serve for one purpose, although their data could be used in multiple manners. Recent developments regarding machine learning and artificial intelligence have opened the door to exploiting historically collected test results by data mining techniques. In an effort to improve the current situation concerning databases, the fib has launched a wide and open initiative on test data management. It consists of a common data structure, sufficiently flexible to serve for multiple purposes, hosting very different test setups, and meant to be used for cross analyses of data. This allows reusing data originally developed for one purpose to analyze other aspects, enhancing the value of current experimental work. Also, a consistent management structure is defined taking advantage of the fib network, with database editors, collectors and users, allowing for a transparent and documented peer‐review process while incorporating new test data. This paper presents the main basis grounding the test data management as well as details on the first two completed applications, referring to fiber‐reinforced concrete (material data) and to punching of slab‐column connections (structural response), showcasing the framework versatility and universality.

Section: A Materials Application Of the Fib Repository: Residual Flex...mentioning

confidence: 99%

Rethinking databases: The fib project for a connected data repository

Fernández-Ordóñez

Ruiz

Tošić

et al. 2023

“…[23][24][25][26][27] ML performs a high accuracy and efficiency in modeling the nonlinear relationship between predictors and outcome variables, and thus it is believed to provide an obvious advantage over the traditional regression approach. 23 Despite a large number of studies using ML to estimate mechanical properties, such as tensile strength, compressive strength, and flexural strength of SHFRCCs, 28,29 very limited research has focused on predicting fracture energy of SHFRCCs in tension.…”

Section: Introductionmentioning

confidence: 99%

Assessment of fracture energy of strain‐hardening fiber‐reinforced cementitious composite using experiment and machine learning technique

Tran

Nguyen

et al. 2022

This study investigates effects of both matrix strength and fiber type on fracture energy of strain‐hardening fiber‐reinforced concrete cementitious composite (SHFRCC) under direct tensile test. Three steel fiber types (twisted, hooked, and smooth fibers) were reinforced to three matrices with different compressive strengths, ranging from 28 to 180 MPa. In addition, a considerable number of direct tensile test results were collected to develop a machine learning‐based model for estimating fracture energy of SHFRCCs. The test results indicated that the fracture energy of SHFRCCs exhibited significant improvements with increasing matrix strength. Moreover, smooth fibers generated the highest values of fracture energy in the matrix with the highest compressive strength of 180 MPa whereas twisted did in other matrices. From the prediction results, the possibility of using a machine learning‐based model to predict the fracture energy of SHFRCCs was demonstrated.

“…They obtained that the coefficient of correlation values of the slump flow, L‐box ratio, V‐funnel, and compressive strength obtained from exponential RBF based SVR model were higher than those of the RBF based SVR model with 0.965, 0.954, 0.979 and 0.9773, respectively. Ikumi et al 26 devised a multilayer perceptron neural network model and predicted the post‐cracking tensile strength of fiber reinforced concrete by using a database including large variety of concrete mixes such as compressive strength, specimen dimension and geometry as well as the fiber content, shape, aspect ratio, tensile strength, and material. Their derived model yielded predictive accuracies of 0.87, 0.86, and 0.83 for the post‐cracking tensile strengths obtained for a vertical displacement of 2, 3 and 4 mm, respectively.…”

Section: Introductionmentioning

confidence: 99%

“…Recently, many researchers [24][25][26][27][28][29][30][31][32] have focused on applying the machine learning (ML) model and deep learning (DL) model to estimate the mechanical properties of concrete. Al-Shamiri et al 24 devised ELM and ANN model to predict the compressive strength of high-strength concrete by using the input variables of cement, aggregates, water, and superplasticizer.…”

Section: Introductionmentioning

confidence: 99%

“…There have been much research [24][25][26][27][28][29][30][31][32] related to the prediction of the mechanical properties of mortar, concrete, SCC. Besides, several studies [33][34][35][36][37] are present about the water absorption of SCC, but according to the literature survey, it is believed that ML and DL based prediction studies about the capillary water absorption of fiberreinforced SCC have been very limited, and there are no published studies yet.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Deep learning and machine learning‐based prediction of capillary water absorption of hybrid fiber reinforced self‐compacting concrete

Kına

Türk

Tanyıldızı

2022

Deep auto‐encoders and long short‐term memory methodology (LSTM) based on deep learning as well as support vector regression (SVR) and k‐nearest neighbors (kNN) based on machine learning models for the capillary water absorption prediction of self‐compacting concrete (SCC) with single and binary, ternary, and quaternary fiber hybridization were developed. A macro and two types of micro steel fibers having different aspect ratios, and PVA fiber were used. One hundred and sixty‐eight specimens produced from 24 mixtures were used in the prediction models. The input was the content of cement, fly ash, silica fume, fine and coarse aggregate, water, superplasticizer (SP), macro and micro steel fibers, PVA, time that the specimen was immersed in water, and splitting tensile strength. Water absorption was used as output. As per the ANOVA analysis of the experiment results, the most effective parameters were macro steel fiber and time for tensile strength and water absorption, respectively. Finally, binary hybridization of 1% macro steel fiber and PVA improved the splitting tensile strength while the use of PVA as binary, ternary, and quaternary fiber hybridization increased the water absorption of SCC specimens. The auto‐encoder, LSTM, SVR, and kNN models predicted the water absorption of fiber reinforced SCC with 99.99%, 99.80%, 94.57%, and 95.50% accuracy, respectively. The performance of deep autoencoder in the estimation of water absorption of fiber‐reinforced SCC was superior to the other prediction models.