Asset health analytics

Data can come from many sources:

• From process instruments. 
• From periodic inspection reports. 
• From continuous monitoring systems. 
• From CMMS systems.
• From manually entered (maintenance) log books.
• ...

Knowing which data to collect is the first step to asset performance management. 

Before modelling and analysis can start, we make data accessible. This can vary from digitizing manually filled out data logs, to collecting periodic data-dumps from hand-held inspection tools, anomaly data transfer and real-time full data transfer.

We can transfer this data to cloud environments or keep it decentralized (“fog computing”).  

Despite its name, big data analytics isn't about throwing massive amounts of data into a computer program, expecting it to spit out novel correlations.

Our specialists can group data into models around failure mechanisms, which will greatly speed up analysis. This includes building algorithms that learn to detect anomalies and failure mechanisms based on historical failures (fingerprints). 

In this next step, we look at what the gathered model tells us about asset performance. This is about determining the reliability of the models, preventing false positives and false negatives, drawing the right conclusions and building business cases to determine the best course of action. 

Visualizing data on dashboards filled with graphs and tables can enhance insights and thereby provide important decision support. We usually see a mix of historic data showing trends, real-time alerts and predictions on the health / chance of specific failures.

Large automation suppliers like IBM, Emerson and OSIsoft provide the software, which we subsequently configure and align with customer-specific KPIs. 

Knowing that something will fail doesn’t immediately dictate what the remedial action will be. We work with our client's teams to review options; including applying some on-line maintenance technique, taking the risk to wait for the next planned turnaround, prematurely shutting down the plant, etc.

In the case of prescriptive maintenance, these decisions are automated, but otherwise humans need to decide. Either way, decisions need to be converted into work orders that will get the right people dispatched with the right tools at the right time to take care of the situation.