Asset Strategy

The “Phenomenological Asset Model” (PAM) is designed to be faithfully reproduce our understanding of “reality”. The PAM is therefore an appropriate tool for developing realistic and optimized asset strategies.

The Phenomenological Asset Model (PAM)

The PAM is a “digital twin” of the asset failure and maintenance processes and their interactions.  The PAM documents our knowledge of each asset vulnerability and the associated maintenance strategy.  For example, Figure 1 characterizes a bearing failure mechanism stochastically, whilst Figure 2 describes the relating Failure-Inspection-Repair processes and their interactions.

Figure 1: Stochastic characterization of (fictitious) bearing failure data.
Figure 2: Schematic depiction of a Failure-Inspection-Repair process and the decision logic for Scenario B.

The stochastic behavior of the PAM and its impact on the production system are estimated via Discrete Event Simulation (DES).

Details of the PAM are not presented here.  However, it is noted that a PAM that faithfully reproduces our understanding of reality may replace alternative approaches such as ACA, RCM, FMEA and RBI.

The importance of modeling decision logic

The PAM must faithfully reproduce our understanding of “reality” to be accepted by the reliability organization.  An important (und underestimated!) aspect that the PAM must address is “decision logic”. For example, Figure 2 implicitly indicates that the bearing is immediately replaced upon detection of the incipient failure.  This logic will have two undesirable outcomes:

  1. unplanned bearing replacement will cause unplanned production loss, and
  2. early replacement prevents utilization of the bearing’s remaining useful life.

In the real world, we apply decision logic to achieve a more optimal outcome.  For example, we may decide to closely monitor the bearing condition to utilize the remaining useful life and “limp” to a planned maintenance opportunity.

Hence, “decision logic” must be considered to develop a “realistic” and “optimal” maintenance strategy.  Interestingly, it is also an aspect of the maintenance strategy that is traditionally neglected.

Three scenarios implementing increasing levels of decision logic are described at Figures 2 and 3.  Each scenario was modeled using a PAM and fictitious data.  The results are summarized at Figure 4.  Details of the models (e.g.: P-F interval, inspection and repair costs, repair times, etc.) are not included here.

Figure 3: Schematic depiction of the decision logic for Scenarios A and C.
Figure 4: The results of PAM modeling of the Scenarios A, B and C (100 simulations, 5-year mission duration).

The average costs for each scenario are presented at Table 1 and demonstrate that the PAM has enabled the specific impact of each scenario and each mitigating measure to be determined in terms of the system performance criteria.

ScenarioLost Revenue Cost ($)Operating Cost ($)Total Cost ($)Total saving
B14726140766 %
C9821631473 %
Table 1: Average costs for the three scenarios (100 simulations, 5-year mission duration).

Notes to Figure 4 and Table 1:

  • Lost revenue cost is the loss of production capacity that results from system unavailability.
  • Operating costs include, in this case, inspection and repair costs.
  • Total savings (Table 1) are calculated in reference to Scenario A.

The above example demonstrates that the PAM is a valuable documentation basis for demonstrating compliance with internal and regulatory requirements.

Unmitigated and mitigated failure probabilities

For compliance and risk-management purposes, it is desireable to estimate the unmitigated and mitigated failure probabilities. These results are able to be extracted from the model results and are shown at Figure 5.

Figure 5: The unmitigated (Case A) and mitigated (Cases B and C) probability of failure for a given time period.

Maintenance data insights

The established consensus in the manufacturing industry is that vast majority of equipment failures are “randomly” distributed, as defined by the Weibull beta parameter:

  • Infant mortality failures: beta < 1
  • Random failures: beta = 1
  • Wear out failures: beta > 1

A knowledge of the beta parameter is useful in determining the appropriate risk management strategy, e.g.:

  • Infant mortality failures: redesign or run-to-failure
  • Random failures: redesign, run-to-failure or condition monitoring
  • Wear out failures: condition monitoring or scheduled replacement

The preconception that failures are randomly distributed, together with the complex nature of the data analysis task in a process plant environment, is perhaps a reason why failure data is often not systematically analyzed.  The consequence is a missed opportunity to optimize the asset strategy.

Weibull analysis of failure data from Case 1 and Case 2 enabled an interesting insight to be gained!

In cases where a Run-To-Failure (RTF) strategy was applied, the resulting maintenance data enabled the correct beta value (1.8) to be determined.  However, the addition of an inspection task transformed the “wear out” Weibull (beta > 1) into a perfect “random” Weibull (beta = 1)!  This is because a random proportion of the failures were prevented via early detection and proactive maintenance. The example illustrates that simulation can help to provide insight into both the effectiveness of the maintenance task and the nature of the resulting maintenance data.  These insights are vital for developing improved Asset Strategies and for performing smarter Asset Analytics based on historical data.

Figure 6: Weibull analysis of maintenace data (failures) for Case 1 (no inspection) and Case B (simple inspection).