PSI - Issue 19

Aaron Stenta et al. / Procedia Structural Integrity 19 (2019) 27–40 Author name / Structural Integrity Procedia 00 (2019) 000–000

39 13

FFS/ASME FFS-1 Part 14 (2007), and (x) given the high consequence of failure for this particular example (e.g. high temperature, high pressure, flammable fluid, and operating in a high population area). Here, failure is defined as crack initiation, however, to expand the network to account for both crack initiation and crack growth, where failure is defined as the limit state when the crack size exceeds some critical crack size, the BDN illustrated in Section 4.2 (Figure 7) can easily be appended to the end of the vibration fatigue network illustrated here. For this particular example, it is assumed that the facility increased the flow rate for increased production, and as a result, increased the piping system of interest’s susceptibility to vibration induced fatigue. To some extent, the operator is aware of the effect of the increased load on the pipe integrity, i.e. increased RMS velocity and stress resulting in reduced fatigue life. However, quantification of the specific expected signal characteristics after the plant load increase are greatly unknown, other than knowing that they got worse. Thus, the facility has to determine whether or not to conduct an in-house measurement to first confirm the signal characteristics. Then, from the measurements, calculate if the piping vibrations exceed the Level 1 allowable RMS velocity vs. effective frequency criteria. If they exceed this criteria, then the piping system is subject to a finite fatigue life. As a result, the site then must decide to either attempt to mitigate the excess vibrations themselves, perform FFS to gain improved guidance on a proper mitigation strategy, or do nothing and continue to operate at the increased plant load. Based on the assumed effect of the increased flow rate on the prior signal, the do-nothing strategy results in high probability of failure after the maximum run time of 25 years. Thus, resulting in a large negative utility of approximately ($65,000,000 USD) at 25 years, so the decision strategy is to stop running immediately. If the facility decides to measure the signal and confirms that the characteristics have increased and fail the Level 1 Criteria, then they have the following options: (i) accept the improved knowledge of the increased signal severity and do nothing (results in a total utility of approximately ($45,000,000 USD) at 25 years, so the decision strategy is to stop running after 15 years), (ii) attempt to mitigate the signal themselves and do not perform FFS (results in a total utility of approximately ($10,000,000 USD), so the decision strategy is to stop running after 20 years), or (iii) perform FFS and then mitigate the signal (results in a total utility of approximately $14,850,000 USD, so the decision strategy is to run for the entire 25 year design life.). Given that the cost of failure in this example is very high, it is obvious that the optimal decision strategy recommended by the network is: Measure? = Yes, Perform FFS? = Yes, Mitigate? = Yes, and Run Time = 25 yrs (maximum time, since the signal was mitigated properly). 5. Conclusions Bayesian decision networks (BDN) are useful tools for recommending optimal decisions at every stage of the lifecycle management of any asset that is subject to any progressive damage mechanism such as fatigue. Specific BDN’s have been developed as part of the ProbFat software tool in order to maximize the average expected annual return (total profit over the lifetime of the asset divided by the lifetime) using a financial cost-benefit approach. The best long-term financial outcome is strived for by maximizing the expected annual return. The management of any asset requires that many decisions be made related to its design, operation, maintenance and repair at different points during its life cycle. In many cases, it is not entirely clear what the best decision is, especially when there are many complex, competing and interdependent factors. For example, as is the case with a coke drum, operating the coke drum more aggressively by decreasing the cycle time will lead to a higher total revenue in a shorter period of time, but this comes at the expense of faster rate of damage accumulation that may shorten the life of the drum and incur more costly repairs and maintenance. However, in many cases, this additional cost is worth it and is more than made up for by the increased revenue. There is usually some intermediate set of operating parameters that provides the perfect balance between these competing forces, and this optimal set of decisions is obtained by using the BDNs presented here. It was shown that, in particular, the optimal cycle time and total time of operation for a coke drum can be found once all costs related to design, repair and failure are accounted for in addition to the rate at which revenue is earned through its normal operation. The optimal decisions were found to strongly depend on these costs and benefits, and this is why any decision strategy must take into consideration these financial concerns in order to lead to the best outcome.