Asset Modelling and Knowledge Capture

Asset Modelling is the structured use of asset analytics, reliability engineering and existing knowledge, to inform decision making.
Asset Modelling and Knowledge Capture

What is asset modelling?

Asset modelling is the end to end process for capturing current understanding (Knowledge) of an asset or asset class for the purpose of decision support. It is the combination of what once was considered separate processes like Failure Mode, Effects and Criticality Analysis (FMECA), condition assessments, probabilistic analyses, task lists, causal modelling and more. Asset modelling combines components of asset analytics, reliability engineering and statistics, with engineering subject matter to be deployed in a composable architecture and at scale.

RCM and Asset Modelling
Asset Modelling vs RCM

Modla's general approach is using a data-interpretation lens, and causal inference.

Now this is different from say, Reliability Centered Maintenance, or machine learning, because asset models contain the causal thinking of Subject matter experts, which make it multi-dimensional and dynamic. When we build the model, we’re not just looking at any one particular asset, but rather at a higher level, of what’s important, why, and how each operational, conditional, and asset specific variable affects each specific asset. This allows you to query the model, and ask the more important "what if" type questions.

What is the aim of asset modelling?

Asset modelling aims to capture Subject Matter Expert (SME) knowledge and the current thinking of industry in a framework that is used to answer many types of business questions (use cases).

Asset models focus on the understanding of an asset, while the analysis, or asset analytics, is an implementation of methodology or business process. Asset analytics concurrently considers the asset information contained in the asset model, along with the operational requirements, risks, as well as the business costs and strategic objectives at fleet or organisational level.

The types of inputs

Asset models can be structured to accept various types of inputs. Each input must impact the asset in some way and usually falls into one of the following categories.

Asset characteristics

Specific asset characteristics impact how an asset can fail. These include configuration attributes, size and type, ratings and build standards. The information typically relates to the modes of failure that are applicable and their initial failure parameters.

Condition data

Condition data determines the current state of the asset. Most analytical approaches assume a new condition. However, up to date condition data facilitates better strategy recommendations reflective of the asset's current condition and age. The conditional inputs are typically measured or observed and can be as simple as age, a visual inspection, or use of more complicated condition monitoring inputs and predictive technologies.

Operational and environmental data

The types of inputs considered under this category, typically reduce or extend the lives of various modes. Operational data includes loading, duty and rate information. It describes how the asset is operated and how its operation affects its performance. Operating attributes can have a direct impact on the component's expected lives.

Environmental conditions include attributes such as process mediums, corrosion rates, or any other factor that can influence asset life, that is external to the asset itself.

Business risk and cost

Business risk and cost data apply the business context to the model. Risk and cost are incorporated using a value framework approach. The approach monetises risk, allowing analytical solutions to maximise value according to the businesses priorities and strategic objectives.

The considered costs also include the usual suspects such as spares and materials, labour, equipment hire etc.

The asset model

The asset model itself may contain any or all of the following elements.

Condition assessment

A condition assessment involves processing conditional data and in turn, making inferences on the condition or health of the asset. The data can be observational e.g. visual inspections or measured e.g. sensor data. Sometimes condition information is used to derive a score or scale that describes the overall health of the asset. However, in the context of asset modelling, condition assessments, age or any other applicable input is used to set the initial failure distributions of the assets. That means the conditional information modifies the statistical likelihoods of failure to account for the current health of the asset.

Sensor data can be analysed by Artificial Intelligence (AI) and Machine Learning (ML) algorithms, but the output is the same i.e. derived conditional information. Predictive Maintenance (PdM) also overlaps with this element. The goal of this analysis in the asset modelling value chain is to establish the links between condition data -> condition assessment -> failure modes.

Failure modes

Stemming from reliability engineering, Failure Mode, and Effects Analyses (FMEA and FMECA variants) form the backbone of our asset modelling approach. By understanding the specific modes of failure that affect asset performance allows us to draw links between how an asset fails and the subsequent events. The asset model structure uses a Reliability Centered Maintenance (RCM) form, comprising of:

  • Components,
  • Functions,
  • Functional Failures, and
  • Failure Modes

Probabilistic analysis

Our probabilistic analysis considers two main aspects:

  1. The probability of asset failure (PoF), and
  2. The probability of that failure resulting in a given consequence (PoC)

Both of these aspects can be affected by environmental, operational, asset-specific, and conditional data. The approach allows us to draw the links between failures, events and the monetary impact on your business. Our platform uses Weibull distributions to represent the probability functions.

Causes

We look at and capture what causes a particular mode or branch of the component tree to be enabled. For example different asset configurations / makes / build qualities.

We look at what causes the failure parameters to change. For example, operating environments, loading or duty.

Inputs

Analysing the causes gives us a good idea of the inputs that the model may require. What variables can we feed with actual data, and what data we may like to collect in the future (Since it has acausal link to the asset performance). We try to keep the inputs generic enough for all users, so some mapping between your data and the model inputs may be required.

These inputs, allow us to add context, and derive a specific asset from this broader “asset model”.

Events

Events are changes of state. Failures, interventions and inspections are all considered events when modelling. Each event can have a cost, risk, resources and tasks. The cost of an event is a combination of the sustained effects and response to that event.

Tasks

Tasks detail what has to be done and when. Tasks are categorised as fault finding or corrective tasks, where fault-finding does not affect the asset in any way, and corrective tasks directly impact the asset's condition and/or performance.

Why do we need to model our assets?

The first step in making better asset decisions, is to understand what asset you have. The second question then becomes, what do you know about them? Asset analytics can answer almost any question that your organisation has, but the more you understand and capture in asset models, the higher quality the results of the analytics piece.

Get in touch!

Want to find out how asset modelling can add structure to your organisations asset knowledge? Contact us here.