Recycled Aggregate Concrete Strength Prediction And Eco-efficiency Evaluation Of LCA-LCC Based On Machine Learning

The application of recycled aggregate concrete（RAC）is important for the construction industry to achieve economic efficiency improvement and sustainable development under the visionary goal of carbon neutrality.As one of the important indicators of the performance of recycled concrete,research on its intelligent prediction based on machine algorithms is still in its infancy.In addition,single studies on the environmental and economic aspects of RAC are predominant,and there is a lack of comprehensive consideration of RAC sustainability in the mix design.In this paper,we obtain 328 sets of RAC measured fit ratio dataset1 based on literature research,and further work on RAC strength prediction and sustainability evaluation by using machine learning method to train the model on dataset1.The main contents are.（1）Machine learning prediction models for RAC strength of reverse transfer neural network,support vector machine and extreme learning machine were established based on the training set of dataset1,and the three machine learning RAC strength prediction models were validated on the test set.The analysis results showed that the support vector machine model was the optimal model,and its fitted decision coefficient R²value,mean absolute error MAE value and mean relative error MSE value were 0.965,1.953 and 14.016,respectively.（2）A multiple output regression prediction model of RAC fit ratio output was established and validated based on dataset1,and its R²value was above 0.95.The data set 2 with different strengths of RAC fit ratios was generated based on the multiple output regression,and then the dataset2 was predicted by the support vector machine model and screened to obtain the dataset3 with the target strength.life cycle assessment（LCA）and life cycle cost（LCC）calculations were performed using the dataset3.The results show that firstly,the environmental and cost optimal fit ratio data for different grades of RAC can be given separately.The environmental impact and cost of the raw material stage of different grades of RAC are much higher than those of the transportation and production stages,among which the raw material stage has the highest environmental impact and cement cost caused by cement production,both reaching about 90%of the total;while for the transportation stage,the environmental impact and cost of natural coarse aggregate is the largest,followed by recycled coarse aggregate and cement;among the six environmental impact subcategories,due to the production of cement and transportation of raw materials in CO₂emissions are larger,resulting in a much larger impact of global warming potential than other impact subcategories such as abiotic consumption potential;the direct costs are much larger than the environmental costs in the total life cycle cost,mainly due to the higher market price of natural aggregates and cement,and the lower unit price that society is willing to pay for environmental emissions.（3）RAC-based LCA and LCC calculations introduce eco-efficiency indicators.Firstly,the data of the optimal eco-efficient mixes for different grades of RAC can be given,where cement and recycled aggregates.Then the correlation of RAC eco-efficiency and strength with parameters such as water,cement,water-cement ratio,and recycled coarse aggregate is quantified using the maximum information correlation（MIC）method,and the results show that the correlation is greater for both cement and water-cement ratio,followed by aggregate.Especially for eco-efficiency,the correlation with recycled aggregates was higher than other natural aggregates.Finally,based on the Sobol index method,we did parameter sensitivity analysis for LCA,LCC and eco-efficiency of RAC respectively,and further established support vector regression prediction models for LCA,LCC and eco-efficiency of RAC respectively,with R²values above 0.9 and good model fitting effect.Then,based on the support vector regression prediction model with high accuracy,sensitivity analysis was performed using quasi-Monte Carlo method,respectively.The results show that LCA,LCC and eco-efficiency have the highest sensitivity to cement,followed by water-cement ratio and aggregates.It is noteworthy that eco-efficiency is second only to cement in terms of sensitivity to recycled coarse aggregates.