Configuring Failure Curve Analytics (FCA) In SAP Asset Performance Management (APM)

[Pardhasaradhi_Reddy_Chelikam]

Have you ever wondered how to predict the Probability Of Failure (PoF) for a technical object associated with a specific failure mode based on its age?. In this blog, I would like to highlight the essential features of Failure Curve Analytics (FCA) in SAP Asset Performance Management (APM).

Failure Curve Analytics (FCA) is a feature within APM that allows the maintenance team to predict when a technical object could fail and how likely  failure would occur given the age of the technical object. The graph that FCA produces is based on Weibull model, a commonly used data model within the realm of reliability. System uses malfunction notifications to train and score the FCA model and visualize the failure data. On the graph, we can view upper and lower confidence intervals (the range in values of likely failure given the technical object`s age) and Probability Of Failure (PoF).

SAP Asset Performance Management

My Business Use Case:

We run an air compressor with a standby equipment continuously to operate the kiln girth gear lubrication system. As a maintenance engineer, I would like to predict the Probability Of Failure (PoF) using Failure Curve Analytics (FCA) in SAP Asset Performance Management (APM) for one of the air compressors before the planned outage in Jul 2025.

My Technical Object Structure in S/4 HANA Asset Management:

MTTR/MTBR Analysis for the technical object 10001729 in S/4 HANA Asset Management:

Before we dive into Failure Curve Analytics, I would also like to touch base on other alternatives to predict next failure date for the technical object 10001729 using MTBF(Mean Time Between Failure) average values. 

• I assume failure occurs at regular intervals and the next failure date is simply:

• I assume failure occurs randomly over time and use random model with exponential distribution to predict next failure date:

Now, let us dive into Failure Curve Analytics and see when the Air Compressor will fail if it runs continuously.

First things first, check the integrity of the Air Compressor 10001729 in SAP APM.

To view the actual failure curve graph, we need to create a FCA model. FCA model needs notifications to train and score. For now, all the notifications & failure modes that will be used to train and score have already been created in the S/4 HANA Asset Management system that is associated with the SAP APM system. 

Create a Technical Object Group and assign the respective technical objects which are being used for FCA model.

Creating Failure Curve Analytics Model:

If we have any model configurations, we can see them once we click on "Configure Failure Curve Analytics".

I have created a new FCA model as,

• The age for conditional probability of failure is used to calculate the conditional probability of the failure. For example, if you enter 100 days as the age and your technical object has a current age of 50 days, then the conditional probability of failure is calculated for the age of 100 days given that the technical object has not failed until the age of 50 days. If you leave blank, then system calculates only the probability of failure.
• First Installation date is used to calculate the age of the technical object at its first failure. If installation date is not maintained, the first failure date is used to calculate the age of the technical object at its next failure.
• If technical object is repairable and restored to a "like-new" condition and set to the age of 0 once it is repaired. This age is used to calculate the age of the technical object at its next failure. If not repairable, the installation date is used to calculate the age of the technical object at its failure.

Under Input Data Sets, let us assign the technical objects that are being trained and scored within the FCA model. We can "Add" or "Remove" the technical objects from the model.

Now, let us assign the failure modes. All the failure modes associated with the failure data profile of the technical object used in FCA model are displayed. We can check off the failure modes to be excluded from training and scoring of the model.

Note: It is considered to have no more than 40 failure modes associated with the model.

Now let us specify the malfunction notification date range to train and score the FCA model. A default date range is provided by the system, although this can be altered if desired.

Training and Scoring the Model:

With all the input Data Sets properly configured, the FCA model AC_SL200_FCA can now be trained and scored.

If the training and scoring fails, system displays the failure list that prevented the model from completing successful training and scoring and the status will be set to "Failed".

If the training and scoring of the model are both successful, the status of the both will be set to "Completed".

If we want to run the model at a different interval, we can still change the interval to DAY, WEEK & MONTH.

Viewing the Analytics Graph of the FCA Model:

Now the FCA Model "AC_SL200_FCA" has been successfully trained and scored. The FCA analytical graph can be viewed within the technical object details page. 

Let us navigate to "Explore Technical Objects" tile.

on the graph, there are three curves. The main one is probability of failure curve, which is the likely percent chance the equipment 10001729 will fail because of the failure mode "Low Discharge Pressure" selected at a specified age in days. The upper and lower confidence intervals mark the highest and lowest percent chance that the equipment 10001729 will fail due to the "Low Discharge Pressure" failure mode. The predicted failure date is Jan 8, 2025, and the time to failure is  47 days.

Key Note:

The MTBF approach to predict next failure date based on average values can mask variations and specific trends in the failure data. They assume constant failure rate, and might not reflect reality.

The FCA with Weibull distribution is more flexible and can represent various failure modes and can model different failure behaviors over time. When we consider systems and their components, they often do not fail at constant rate, So the FCA analysis can account for varying failure rates over time and fits the actual failure data to a curve and predicts when future failures might occur, where as the MTBF cannot. 

This brings me to the end of my process walk-through. I hope this blog proves useful to Plant Maintenance team. I look forward to seeing Plant Maintenance team using SAP APM Failure Curve Analytics to predict the technical object failure. 

If you would like to keep up on SAP Sustainability Control Tower, Asset Management and Asset Performance Management, follow me on SAP Community and LinkedIn.

--

Yours Sincerely,

Pardhasaradhi Reddy.C

My LinkedIn Profile: www.linkedin.com/in/pardhreddyc