Prefabricated Bridge Beam Quality Evaluation Model Based on Machine Learning:Taking Beam Concrete Strength for an Example
To address the issues of incomplete evaluation criteria,ineffective reflection of quality differences,and lack of process control indicators in the current quality evaluation on prefabricated highway bridge beams,this paper proposed the quality evaluation method for bridge beam prefabrication process based on the machine learning.IoT data from key process controls(e.g.,concrete test blocks,concrete mixing,pouring and vibration,rebar testing,beam steam curing,and intelligent tensioning and grouting)were collected.The collected data were normalized,then filtering out anomalies and outliers,and achieving data denoising.Pearson correlation analysis algorithm was used to establish the correlation analysis matrix diagrams.The strongly correlated factors influencing beam quality by key process control deviations were identified.The data dimensionality reduction was achieved,providing the high-quality denoised data for machine learning.The various machine learning algorithms(e.g.,linear regression,decision tree regression,and XGBoost)were used to establish the beam quality evaluation models with the processed data.The established models were quantitatively evaluated by using RMSE and MAPE.The optimal beam quality evaluation model was identified for quality evaluation.The beam concrete strength was taken for an example.The test result indicates that XGBoost model achieves MAPE of 1.87%in predicting the beam concrete strength,significantly lower than the 3.59%error rate of the rebound method currently used in the industry.The proposed method not only accurately evaluates the concrete strength quality of finished beams,but also identifies the key processes and indicators affecting the beam prefabrication quality.That enables the precise control on these key links.Furthermore,it can effectively and accurately detect the instances of human fraud in beam quality testing,and promote the overall homogenization and improvement of prefabricated beam quality.
NETL
NSTL
万方数据
