You are Home   »   News   »   View Article

Analytics driven predictive asset management

Friday, September 13, 2013

It is time that the oil and gas industry moved from predictive maintenance -doing maintenance tasks according to a predefined schedule - to an analytics driven approach, where you make better predictions of what maintenance needs to be done, said Deb Chakraborty, Associate Partner, Asset Management Solutions at IBM

Most offshore assets are still maintained according to ''preventative maintenance'' systems, where maintenance is done according to a schedule.

''Preventive maintenance, absolutely it''s not optimum,'' said Deb Chakraborty, Associate Partner, Asset Management Solutions at IBM.

''The preventive maintenance is usually based on the calendar maintenance schedule.''

He was speaking at the Digital Energy Journal conference in March 2013 in Aberdeen, ''Improving offshore facilities management, operations and maintenance''.

Preventive maintenance schedules are ''usually expressed by the manufacturer, but manufacturers only have a very generic accommodation of the maintenance requirement of the asset.''

For example, there might be a much faster deterioration in a pump which is pumping seawater offshore, compared to if the same pump was pumping demineralized water on a land power station.

And the pump might need a lot less maintenance when it is 5 years old compared to when it is 30 years old.

Preventative maintenance is the equivalent to driving a car looking at of the rear view mirror, he said. We need a way of planning maintenance where we are looking through the car windscreen.

Analytics prediction

''We invested some of our time in developing a smarter asset management solution which starts from analytics predicting, using asset management data, predicting the future of assets.

''I believe that all the asset management community should be smart in moving towards that''.

Asset management systems are gradually moving from ''computerized maintenance management systems'' to building information modelling systems (BIM).

But asset management is still considered a kind of maintenance activity, he said.


The first thing which is needed to do analytics is data.

The biggest source of suitable data is in systems like Maximo and SAP, he said. There is a large amount of data available in these systems.

But there are many other sources of data which can be helpful understanding current asset health and predicting future performance.

For example telemetry /sensor data and control system data can provide a clear picture of asset performance in the past, helping make a prediction of future performance, he said.

Also you can use the design data. ''There''s huge amount of data that is generated and stored in the engineering information system, during the design, during the construction stage,'' he said.

There are often problems integrating design data at the plant hand over stage, when the information is handed over to the company which will own and operate the plant. A lot of processes end up being manual, and a lot of information gets lost.

Analysing data

To analyse the data, first of all these different data stores need to be integrated together. You don''t need to necessarily change the way the data is being stored - but you can make a copy of this data in another database.

To work on it, you can use sophisticated tools, such as IBM''s ''SPSS Statistics'' (originally called Statistical Package for the Social Sciences) to build a deterioration model or risk profile, he said.

You can identify the probability of an asset''s failure at a certain point in the future. ''Of course confidence in the output depends on the quality and availability of the data,'' he said.

One way of making up for missing data is to use text analytics. Good amount of rich information are usually captured in operators'' log book. SPSS can do text mining to uncover useful insight from those unformatted texts.

In one of text analytics engagements it was found that the failure of the plant is also more likely if there is more than two corrective work orders were due. ''We identified that if there are two or more corrective workorders, overdue the failure is about 96% of time,'' he said. ''When there is one less overdue, the failure went down to 78%. ''This is a very interesting insight.


Often, users complain that the reporting they get from systems is too complicated to work with, he said.

IBM''s solution was to use its ''Maximo'' asset management software and put all of the data in the Maximo database.

''We created a custom table, integrated all that to an SPSS server and brought back the result of that analysis to Maximo,'' he said.

Mr Chakraborty also mentioned that data captured in other plan engineering software like Bentley Nevada vibration monitoring system could be used in the failure modelling.

The end result is a system where users can see quickly what is happening to their asset, what is expected in the future, all on one system. It can be done as part of their usual work.

Building on this system is a traffic light model for the whole plant, where an offshore platform manager can see which components on his platform have the highest amount of risk at any time. ''We''ve divided it into green, amber and red based on the probability of the asset''s failure at certain point of time.''

Then you can look in more detail at the ''red'' items and find out more about the risk - including the probability of failure at a certain time, and the accuracy of the prediction. You can see when the component failed in the past and why.

Then you make an estimation of when the maintenance should be done, and create a work order.

''So a very complex statistical analysis suddenly becomes very user friendly and actionable,'' he said.


Mr Chakraborty gave some examples of companies which he had worked with, to improve asset management.

One oil company achieved a saving of $42m using the solution, he said.

One helicopter manufacturing company claimed that it reduced its direct maintenance cost reduced by 25% with the system.

AUK water utility, used the system to analyse its sewage blockage. The analysis identified that a particular sewer was suffering from a huge amount of blockage - and it was then able to identify a manufacturing company nearby which was using the sewer as a rubbish dump.

Associated Companies
comments powered by Disqus


To attend our free events, receive our newsletter, and receive the free colour Digital Energy Journal.


Latest Edition Oct-Dec 2023
Nov 2023

Download latest and back issues


Learn more about supporting Digital Energy Journal