Review of the Bureau of Reclamation's Corrosion Prevention Standards for Ductile Iron Pipe (2009)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Appendix B Dissenting Statement This statement concerns the interpretation of the evidence of failures in duc- tile iron pipe (DIP) with polyethylene encasement (PE) and cathodic protection (CP), and of the interpretation of field data, for which the author, a member of the Committee on the Review of the Bureau of Reclamation’s Corrosion Prevention Standards for Ductile Iron Pipe, has different views from those stated in the body of the report. Relevant issues are noted below. The first issue concerns the number of failures to merit consideration as evi- dence of the corrosion performance of DIP with PE and CP. That number was chosen to be three in the analysis presented in Chapter 4, “Failure Criteria,” cor- responding to the Southwest, Akron, and California City entries in Table 4-2. The author differs by noting that two of those failures have serious disqualifying fac- tors. As detailed in Chapter 3, the California City event took place in an extremely aggressive environment (100 ohm-cm) where the pipeline had served half of its pre-failure life without any cathodic protection. Such a situation falls far short of that of a properly maintained and operated cathodic protection system, and the author deems it not relevant to the conditions of interest. Also as indicated in Chapter 3, the Akron, Ohio, failure occurred at a spot where the presence of inter- segment electrical bond, necessary for proper cathodic protection, could not be confirmed. As in the previous case, authentication of the failure as representative of regular operations cannot be supported by the author. It is also remarkable that both failures took place in very short segments of pipe (1.5 miles and 0.25 mile for the California City and Akron failures, respectively), which were also the two short- est pipe segments listed in Table 4-2, and each far shorter than the length-weighted 151 

152 Corrosion Prevention Standards for Ductile Iron Pipe average of all the entries. Those observations further suggest a singular character for those failures, possibly related to the level of maintenance, inspection, and control resources that may be available for large as opposed to small projects. The only remaining failure (Southwest) does not appear to have major potential disqualify- ing issues except for reported difficulties in consistently meeting CP goals as well as potential PE damage from aggregate during placement. Consequently, the author estimates that only one failure merits serious consideration for analysis against the expectations from the benchmark stated by the Bureau of Reclamation. The second issue concerns this author’s disagreement with the analysis methodology used in Chapter 4 to compare evidence of field failures with the expectations from the benchmark of 0.000044 failures per mile per year stated by Reclamation, which will be considered as an agency-specified parameter in the following. The author contends that because of the very sparse DIP with PE and CP failure data set, interpretation of those data to calculate a nominal fail- ure frequency for comparison to that of the benchmark is not appropriate, as it is akin to comparing over a short period of time the nominal death rate from a small community to that of a large city. Thus, comparison between this nominal rate and benchmark rates, as used in Chapter 4 and cited in Chapter 6, “Findings, Conclusions, and Recommendations,” is not warranted in the view of the author. The corresponding sensitivity analysis in Chapter 4 of the report does not resolve this concern, as such analysis would be an extension of assigning undue signifi- cance to the nominal rate. The author proposes instead to estimate, using the benchmark rate, the prob- ability of having the number of failures observed in the DIP with PE and CP experi- ence inventory for the given amount of pipe length-years in that inventory. If that probability is found to be appreciably large, then the DIP with PE and CP failure data set may not be indicative of diminished performance compared to that of the benchmark set. Conversely, if that probability is found to be very small, the DIP with PE and CP failure data may be seen as indicative of diminished performance relative to the benchmark. It is emphasized that this analysis is limited only to the implications of observed failure events. Other evidence of performance such as the presence of corrosion in the absence of failures was considered elsewhere in the report, as well as in discussions based on corrosion engineering principles. In the following, the benchmark failure rate stated by Reclamation will be assumed to be numerically equal to the probability P of one failure occurring per mile per year in a hypothetical large reference system (or “benchmark system”), so P = 4.44 × 10–5. That assumption is adopted considering that the arguments presented by Reclamation in developing the benchmark involved a considerable  Graham E.C. Bell, Schiff Associates, “Measurements of Performance of Corrosion Control Mecha- nisms on DIP,” presentation to the committee, Washington, D.C., July 29, 2008. 

A pp e n d i x B 153 number of failures (on the order of 20) occurring over several years on a very large pipeline length, and that the corresponding number is <<1. The above consider- ations are comparable to those used in Chapter 4, in what pertains to treating the numeric Reclamation benchmark as a probability. The present analysis takes a different direction from here on. Assuming that the annual probability of failure per mile P remains constant over time and that each mile of pipe and each year is statistically independent from the rest, the probability P0 of having zero failures in the benchmark system in a given T year period for a given length of pipe of L miles can then be determined as follows: P0 (%) = 100 (1 – P)TL ~ 100 e–PTL (TL >> 1). (Eq. B-1) Thus the probability P1+ of having at least one failure in the benchmark system over a given time and length interval is estimated by P1+ (%) = 100 (1 – e–PTL) (TL >> 1). (Eq. B-2) Following a similar analysis the probability Pn+ of having at least n failures in the benchmark system for a given length and time interval is estimated, when TL is large, by Pn+ (%) = P(n–1)+ – 100 (PTL)n–1 e–PTL /(n – 1)! (Eq. B-3) which is a form of the Poisson distribution. Table 4-2 shows that the DIP with PE and CP documented experience concerns an integrated interval TL ≈ 7.94 × 103 mile per year. Consequently, by application of the above equations the probability of having at least one failure in that same TL interval in the benchmark system would be approximately 29 percent. For at least two failures the probability drops to about 5 percent, while for at least three failures it is only about 0.5 percent. The import of this finding together with the discussion on the first issue becomes apparent. With only one qualified failure, the probability of having at least one such failure in the benchmark system for the same integrated time-length is quite high, ~29 percent. Thus the single qualified failure could be easily dismissed as an event that would have been frequently observed in other surveys of comparable pipe length and duration in the benchmark system. On the other hand, if even two of the failures had merited qualification, the probability would have dropped to the much smaller ~5 percent value, while if all three of the failures considered  J.H. Pollard, A Handbook of Numerical and Statistical Techniques (Cambridge: Cambridge Uni- versity Press, 1977). 

154 Corrosion Prevention Standards for Ductile Iron Pipe were pertinent, the probability of their having been observed in the length-time interval considered for the benchmark set would have been exceedingly small. The author proposes therefore that, absent solid evidence for more than one qualified failure in the ≈7.94 × 103 mile per year experience base, the failure data alone do not sufficiently support stating that DIP with PE and CP has clearly poorer corro- sion performance than that of the Reclamation benchmark system. The third issue concerns the significance assigned in parts of the report to linearized corrosion rates values calculated from field and yard test measurements of corrosion penetration. Such values were largely considered outside the context of how frequently along the pipe and over what time period the sampling took place. In field samplings, localized penetration data are usually obtained at points where signs of severe corrosion distress had been observed, and often because a failure was noted. Therefore, it is not surprising that localized penetration rates there are high, since whichever protection system was present had clearly failed at that location. In the author’s view, the main concern in these rarely occurring events should not be about what the rate of localized corrosion progression was, but instead about what the frequency of those events per pipe length-duration was, as in the focus of the previous two issues. Hence the meaning of those calculated values, at least as far as they concern field performance, is deemed by the author to be too limited to support or oppose the conclusions in the report derived from other considerations. In summary, the author shares the committee view that the scientific under- standing of corrosion mechanisms casts serious doubt that DIP with PE and CP can guarantee a long service life in highly corrosive soils. However, the author does not agree that the available failure data and calculations of linearized corrosion rates from field measurements provide conclusive enough supporting evidence for that view. Alberto A. Sagüés, Member Committee on the Review of the Bureau of Reclamation’s Corrosion Prevention Stan- dards for Ductile Iron Pipe  Theauthor notes that the committee considered two alternative benchmarks, one close to the Reclamation value and the other about 4 times smaller. The calculated probability of at least one failure for the latter is about 9 percent, which would suggest some concern if even one failure were qualified. However, potentially much less conservative alternative benchmarks (with consequently much greater probability of a single failure) are possible as well if correcting steel pipeline data for the fraction of soils that are highly corrosive. Such alternatives were recognized but not quantified by the committee in the absence of suitable data, but that absence if anything further emphasizes the inherent uncertainty in assigning undue significance to the presence or not of just one failure.