Asset Modelling is the end to end process for taking asset data to decision making. It is the combination of what once was considered separate processes of FMECA, Condition Assessment, Probabilistic Analysis and Optimization Techniques and more, into a single, integrated approach.
Aims of Asset Modelling
The aim of modelling an asset is to make informed decisions around:
All while considering operational requirements, environment, risk, and business goals.
Condition data is used to determine the current state of the asset. Most analytical approaches assume new condition, however knowing the exact condition of the asset allows intervention recommendations to be made as well as evolving strategies to be adopted. These inputs can be as simple as age, or visual inspection results, to more complicated condition monitoring inputs using sensors.
In order to understand how the asset can fail, the specifics of the asset in question must be known. This data set includes configuration specifics, size and type, ratings and build standards information.
Operational data includes loading, duty and rate type information. How the asset is operated can have a direct impact on the expected life.
Environmental conditions include process mediums, corrosion areas, or any other factors that can influence asset life, outside the asset itself.
Risk assessment for failure, operational, intervention and maintenance events.
Cost for components, labor, equipment hire etc, as well as risk cost.
Condition assessment is the process of taking the condition data, understanding it, and deriving conditional information. The information can be used to "Score" the health of the asset, but also (and less common) to drive the initial conditions of the failure analysis. AI and Machine Learning can be used in this space to analyse sensor data. Predictive Maintenance (PdM) also overlaps with this element. The goal of this analysis in the Asset Modelling value chain, is to link condition assessment outputs to the failure mode inputs.
Stemming from Reliability Engineering, Failure Mode Analysis (FMEA and FMECA variants) is the backbone of most asset modelling approaches. With AI and Machine learning becoming increasingly popular, there are other structures that are used here.
There are 2 main parts to the probabilistic analysis that need to be considered. The Probability of Failure (PoF) and Probability of Consequence (PoC) both of which can be affected by the environmental, operational, asset specific, and conditional data. Understanding these 2 components allows the links between failure and cost to business to be established.
Events are changes of state. Failure states, Interventions, Routine Maintenance etc. are all considered Events when modelling. Each event can have a cost, risk, resources and task.
Tasks detail what is to be done to the asset and when.
Maintenance Strategy optimization works by varying the applicable tasks and timings to determine the combination that delivers the most value, for the lowest cost. Maintenance strategy optimization includes consideration of Risk Based Inspections (RBI) as well as adaptive and online strategies.
Operational strategy optimization is less common than Maintenance strategy optimization, however it still has it's place in increasing value. It works by altering the operational requirements of the asset, and measuring the trade-off in terms of reduced asset life. Overloading or under-utilization can have a dramatic impact on asset performance as well as life. This optimization approach aims to determine the operating parameters that will deliver the most value over the assets life.
Typically capital works. Interventions include replacements, and major refurbishments.
The complete value chain from condition to tasks, when coupled with costs and risks, can be used to generate profiles for budget setting and reporting.
Equipment selection is an important decision making point for replacement equipment as well as new or greenfield projects. Optimum equipment selection is determined by varying the Asset Characteristics to select the best combination for the environment and operational requirements.
Reliability Engineering tip of the week: As a former control systems engineer, historian (e.g. PI) data is a great first place to look when conducting RCA’s or Investigations. Use trends and log files to help establish an evidence-based timeline or sequence of events.There are too many times when I’ve had an operator say that “We didn’t push the stop button”, when the control log says the stop command was initiated from their workstations!
Condition scores are typically an output from a condition assessment. Using only condition scores for decision making is not recommended. The primary problem with this approach is that it does not consider risk and/or value to the business. As an example, consider a typical power pole from an Electricity Distribution Utility.
1) A Poor condition pole in the middle of a desert serving a single home.
2) A Poor but slightly better condition pole in the city serving 100 customers with a much larger safety risk and reputation impact.Which would you replace first?
While they can help us derive insights from the data, these insights need to be understood before handing over control to the algorithm. I believe that the best approach is to balance the AI insights with engineering principles (reliability engineering in an asset space), validating the analysis of the AI, and linking it back to something explained and understood. If you don’t then:
1) You don’t learn as much,
2) You are exposed to issues / faults in the data,
3) It's harder to correct and improve the algorithms.