An innovative method for estimating the compressible strength of concrete is suggested in this study and is dependent on the machine-learning technology. This method uses the dynamic boosted technique to combine numerous weak classifiers into a strong classifier that can identify the connection between input and values. The total reliability of the strong classifier will be improved since the weak classifier with the tiny forecasting mistake will be given more value in the network. To train and assess the learners, a maximum of 1030 frames of tangible compression tests have been gathered, where the input signal are the elements of the cementitious material (such as coarse/fine agglomerates, concretes, moisture, thickeners, etc.) and the healing time, and the data flow are the fracture toughness values. The suggested technique produces an average reliability of above 95% in notion of coefficient of determination ( r2) and is verified using a ten-fold cross validation procedure. In order to show the suggested mode's generalizability, a fresh collection of 103 trials for compressive strength in concrete is also employed. Artificial neural network (ANN) and support vector machine (SVM) are two additional separate machine learning techniques that are currently being used in this sector, and the suggested methodology outperforms existing approaches in every way. The impact of various important adaptive boosting technique components, including as the quantity of training examples, the selection of base learners, and the impact of the sensitivities and variety of input characteristics, is also examined in the last section. It is demonstrated that the logistic regression is the best option for the generalization error in the boost architecture when utilising 80% of the complete data as training data. Additionally, based on the findings of the parameter estimation, the significance of various input factors is determined.
Key words: Machine , Compressive , Strength , Civil , Structures
|