Applying Data Mining Technique to Compute LDE for Rutting Through Full Scale Accelerated Pavement Testing

摘要

Results from closely controlled full-scale Accelerated Pavement Testing (APT) were employed to establish a rutting prediction model using the "Find Laws " data mining technique. Seven test pavements (264 records) from the Cold Regions Research and Engineering Laboratory's (CRREL 's) Heavy Vehicle Simulator (HVS) and one test pavement including 8 records from the Texas Department of Transportation's (TxDOT's) Mobile Load Simulator (MLS) were included in the model development. For model verification purposes, additional test pavements and 35 records from CRREL's HVS were utilized. Find Laws was applied successfully to develop a rutting prediction equation that uses wheel load, load repetitions and the pavement Structural Number (SN) as inputs. CRREL's HVS test pavements had wheel loads of 20kN to 103.5kN, while TxDOT's MLS test pavement had only one wheel load of37.8kN. The overall R~2 for the 272 records is 0.6745. To compute the Load Damage Exponents (LDE), rutting for SN from 2 to 7 under wheel loads from 20kN to 90kN were computed using the developed model. LDEs were computed for each SN and were found to be reducing with increasing SN values. LDE values are 7.58, 5.62, 4.68, 4.11, 3.71, and 3.41 for SN values of 2, 3, 4, 5, 6, and 7, respectively. It demonstrates that overload has more pronounce effects on rutting performance for weaker pavement (lower SN). For example, for a 20% higher load there will be approximately 400% higher rutting when the SN is equal to 2. But it will be only 197% higher in rutting when the SN is equal to 7. The model and algorithms proposed in this study provide a good foundation for further refinement when additional data is available.