134

It is widely accepted that the yield strength of low carbon and low alloy steels can be modeled as the sum of contributions from intrinsic lattice friction stress, solid solution strengthening, grain size effects, dispersion strengthening, and dislocation density. Dispersion strengthening can be neglected for hot-rolled steels that aren’t microalloyed with Nb, V, or Ti, while models for solid solution (Pickering), grain size (Hall-Petch), and dislocation density (Taylor) effects have been extensively validated in published research dating back as far as the 1930s. These models have been shown to be accurate and effective, but line pipe applications have been limited because the quantitative metallography required to determine grain size has historically been time consuming and resource intensive. The recent development of a machine learning model to perform rapid, reliable grain size measurements warrants revisiting these models.

In the current work, the yield strengths for a set of 39 hot-rolled, non-microalloyed steels are estimated from the %Mn, %Si, %Cr, %Cu, and grain size using the model published by Pickering in 1978.[1] The model calculates the grain size and solid solution contributions using only the original coefficients published in 1978, while the contribution of dislocation density is estimated by assuming a constant, average dislocation density for all hot rolled, non-microalloyed pipes. Comparison of model predictions to yield strengths from tensile testing resulted in a mean absolute percent error (MAPE) of 8.0% with a 95% prediction interval of ±11.8 ksi. This performance compares favorably to the instrumented indentation test (IIT), where comparison to tensile testing for an expanded set of 83 pipes yields a mean absolute percent error (MAPE) of 8.6% with a 95% prediction interval of ±12.2 ksi. In a direct comparison of 30 pipes for which both sets of data were available, yield strengths estimated by the model and by IIT predicted the results from tensile testing with MAPE values of 8.2% and 6.8%, respectively, and 95% prediction intervals of 12.0 and 12.5 ksi. The results suggest that this model may be a useful tool to confirm strength testing by IIT, provide strength estimates where IIT is not practical, or replace some fraction of strength testing altogether.

[1] Pickering, F. B. Physical Metallurgy and the Design of Steels; Applied Science Publishers: London, UK, 1978; p. 63.

Read/Download PaperGradeAvailable after the event