3. ERRORS BY GTI
Throughout the GTI report, the work done and the analysis presented multiple errors that resulted in misleading figures and conclusions regarding the HSD technology and material strength determination. We’ll describe a selection of these errors in further detail below.
3.1 GTI Incorrectly Compared Datasets
PRCI blind-tested alternative nondestructive material verifications and published report NDE 4-8 . PRCI found that the HSD, compared to Frontics and other NDE techniques, is “the best performing technique.” Table 1 depicts the conservative shift for multiple processes below.
Surface vs. Bulk
NDE material verification processes perform a direct measurement near the surface on the outside diameter of the pipe to infer the bulk properties for the entire pipe wall. For welded samples, surface strength is generally higher than the bulk values determined through destructive lab testing. This difference results from manufacturing where the material is often cold-rolled and bent for a pipe.
The GTI report uses “base” or “bulk” when referencing the destructive values or the nondestructive measurements that correspond to destructive values .
How Each Process Works
While both technologies provide estimates of bulk strength values offered commercially, the methods used by MMT and Frontics are markedly different, as shown in Figure 2.
MMT uses a contact mechanics technique known as frictional sliding to gather raw surface data, which is input into a Finite Element Analysis (FEA) simulation to generate a surface strength value. The surface strength value is an intermediate process variable. Additional field data, including chemistry and microstructure data, is used in MMT’s machine learning model to deliver the final strength estimate provided as the commercial offering.
Frontics uses compressive indentation testing results, which are used directly in their machine learning for
final strength estimates provided as the commercial offering. There is no physical modeling using FEA within
the Frontics process. Without this step, there is no surface yield strength. The raw data is provided to machine learning models for final strength determination. Specific vendors may opt to use additional field data and adjust the final value. Still, the use of other field data will vary from vendor to vendor.
What GTI Did
GTI requested MMT to provide the intermediate process surface strength variable, which is not provided as part of the commercial MMT process. The intermediate surface strength from the MMT process was used as a part of the GTI analysis and compared to the Frontics final bulk values in GTI report Figure 12, where GTI incorrectly labeled Frontics results as surface values. As verification that GTI should not have labeled Frontics results as surface strength, Table 7 of the GTI report depicts the master database and show no variable corresponding to a Frontics value for surface strength .
Due to the GTI report and the incorrect comparison of datasets, operators can be misled into a faulty understanding of tool performance. One example of this misleading representation is shown in Figure 3, which is produced and corrected.
(A): GTI report figure 11 depicts an example of misinterpreted data, comparing “Frontics Surface,” which is mislabeled and should read “Frontics Bulk,” and the MMT Surface data to lab values.
(B): Shows a corrected version of GTI Figure 11, correctly comparing NDE bulk measurements to laboratory values.