Evaluation of Pile Capacity Prediction Methods Based on Cone Penetration Testing Using Results from I-15 Load Tests

Abstract

Nine axial pile load tests were performed in connection with the reconstruction of I-15 through Salt Lake Valley, Utah. In addition, load test data were collected for seven other piles in the valley. The soil profile at these test sites generally consisted of saturated soft to medium stiff silts and clays underlain by a dense sand bearing stratum at 25 to 30 m. Test piles consisted of closed-end steel pipe piles 300 to 600 mm in diameter with a concrete fill. Because cone penetration soundings were available for 15 of these test piles, these data were used to evaluate the accuracy of five available design methods that are based on cone penetration testing. Methods evaluated were the Dutch, Schmertmann, Tumay, Laboratoire des Ponts et Chaussées (LPC), and Fellenius methods. The Davisson method was used to define ultimate capacity. For most design methods, the estimated skin friction was on average only 40 to 60 percent of the measured resistance. The Dutch method predicted 82 percent of measured skin friction, on average, but still exhibited significant scatter for individual tests. Underprediction of skin friction by the LPC method was caused by unnecessarily severe restrictions for steel piles. Without these restrictions, the standard error was cut in half and estimated capacity was 96 percent of the measured capacity, on average. The LPC method also provided the best match with measured end bearing resistance (within 11 percent on average), whereas other methods overestimated capacity by 60 to 80 percent. Overall, the LPC method without side friction limitations gave 97 percent of the measured capacity on average and had the lowest standard error of any of the methods.