PSI - Issue 19

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

38 12

vibration limits for fatigue susceptibility. In vibration applications we are interested in cyclic failures in the very high cycle regime, i.e. out to 10 7 cycles and beyond, unlike the low-cycle fatigue of a coke drum that shows signs of crack initiation in as few as a hundred cycles. In other words, the stress range (vibration magnitude) is small but the number of cycles (frequency) is very large. The BDN illustrated in Figure 8 has been designed to illustrate a common optimal decision strategy for flow induced vibration fatigue of mainline piping subject to random vibrations. This paper utilizes the Level 1 (L1) FFS criteria proposed in Bifano (2018). The decision nodes to be optimized during the Single Policy Update are Flow Change?, Measure?, Perform FFS?, Mitigate?, and Run Time?. The determination of the optimal decision strategy depends on the cost/benefit that each decision provides. Here, the cost/benefit of each decision is assumed relative, i.e. Measure Cost = ($1,000 USD), Mitigate Cost = ($50,000 USD), FFS Cost = ($100,000 USD), Failure Cost = ($100,000,000 USD), and Run Benefit Per Year = $1,000,000 USD. Note, as illustrated in Figure 8, each Signal node is a group node that represents a collection of 3 nodes that characterize the signal. Node groups are used for illustrative convenience only, and do not impact the outcome of the network.

Fig. 8. Vibration Very-High Cycle Fatigue Life-Cycle Bayesian Decision Network.

In this representative network, the following assumptions are made: (i) from a fatigue perspective, the signal can be characterized by the combination of the RMS velocity, frequency, and kurtosis, (ii) the geometry, design, and installation have already been chosen (here, we assumed the circumferential weld of a mainline subject to random vibrations from an increased flow rate), (iii) if a measurement is performed, it is done by the facility’s in-house operator, who then performs the simple L1 screening calculation, (iv) measurement error for the in-house operator’s measurement is assumed high as the operator is not likely to identify the location of maximum vibration, (v) if the piping component fails the L1 criteria, then the client has the option to mitigate, perform FFS only, or perform FFS and mitigate the signal, (vi) using primarily the knowledge from the measurement, the facility may attempt to mitigate the signal without performing FFS, however, the ability to select the proper location and design for mitigation is quite low and ineffective, (vii) as a result, the mitigated signal is assumed to have less variance if FFS is performed, (viii) after mitigating the signal the L1 method for the mean cycles to failure is assumed, (viv) the fatigue curve is the equivalent structural stress that is described in detail in Level 2 Method C of API 579-1