Updated: Nov 27, 2019
Are you making the most of new technology?
As a plant operator, Unplanned Downtime (UPD) can be the bane of your existence. Just for reference, I am referring to any unplanned event that stops production for any amount of time. Unplanned downtime is most often caused by operator error, poor maintenance, hardware or software error and ‘perceived downtime’, which includes poor performance or slow changeovers.
Using the offshore O&G sector as an example, a recent report by Baker Hughes states that “Offshore oil and gas organizations experience on average $38 million annually in financial impacts due to unplanned downtime. For the worst performers the negative financial impact can be upwards of $88 million pa.” In other energy sectors, reports show that the average annual cost for a midsized LNG facility relating to UPD is $150MM and the list goes on.
No one argues that UPD is a major issue but the solution to minimizing this can be linked to your maintenance strategy. The three key implementations we at Digital Plant Specialists (DPS) are seeing in the market are Reactive, Preventative and more recently Predictive.
So how does each strategy work?
This system, aka “run to failure”, is a maintenance strategy where work is carried out only after a breakdown or failure occurs. The main reasons for running with this are that it is very simple to understand and use, it requires minimal effort to manage and operate and can be used without installing pricey software and sensors.
Whilst this means a lower up-front cost to the plant owner, the downsides are potentially quite significant. There is a high probability of UPD which in turn will reduce life expectancy of plant equipment and result in a very unsafe working environment. A significant number of major LTIs and fatalities can be traced back to equipment failure. In 2004, an explosion at the LNG plant in Skikda, Algeria, resulted in 27 people killed, 72 injured, and seven reported missing. The explosion destroyed three out of six liquefaction trains, damaged a nearby power plant, and led to the shutdown of a 335,000 bbl per day refinery. The suggested causes included “poor maintenance and poor general condition of a boiler unit” 1. According to the Marsh report “The 100 Largest Losses 1978-2017 – 25th Edition” the adjusted cost of the explosion was estimated to be $689MM.
Can an company in today’s operating environment afford to take the risk of such failures happening due to an outdated maintenance methodology?
This strategy is commonplace in many plants operating today. It relies on a routine maintenance schedule to keep the equipment up and running. The maintenance is time based and performed at set intervals which are derived from average equipment life, wear and other information available from manufacturers and in approved data bases. The benefits of such a system are numerous and when implemented with a Computerized Maintenance Management System and associated condition monitoring sensors will give you a better insight into your maintenance activities and improve the life span and reliability of your plant. However, such a system does not generally take into account asset wear and process changes (e.g. an increase in production rates for a period of time due to demand changes) which may be short term but will take the equipment involved outside its integrity operating window hence decreasing its expected life. In addition, alerts raised when critical equipment is logged as being High or Low based on the parameters set by the operations team, will only let you know that there is a problem to fix now, a better scenario than waiting for fail but still results in a shutdown that whilst manageable, is still unplanned.
The ultimate solution, available and practical today, is Predictive Maintenance. This is where your solution is predicting when a piece of equipment will fail so that maintenance work can be planned and performed just before it happens. The predictions are based on the real/near time data gathered by a smart sensoring system and combined with the historical or legacy static data gathered over the plants life.
When the system is operational, and this will take some thought and time, predictive maintenance has the ability to not only give you a real time view of the current conditions of your plant but will ensure that there are minimal disruptions to operations, maximising asset up time. Consequently, maintenance operations and spare parts inventory management become optimised, enabling a safer working environment, cost reductions and revenue enhancements.
Predictive Maintenance does come at a cost which hits both CAPEX and OPEX and will take time to set up. Machine Learning Algorithms, the backbone of the software, by name, require time to learn, on clean data sets to be able to predict issues related to the specific plant and it also requires training in new technologies for the company as a whole. Nevertheless, companies that are heading down this digital plant solution, are already seeing ROI and significant decreases in safety incidents. Vroc AI, one of the leading Predictive Analytics companies in the Energy and Utilities space have stated that in the past year, their clients have seen over $30MM in cost savings and, more importantly, over $100MM in production loss avoidance.
The Digital Transformation Journey, from where companies are now to implementing and seeing the tangible benefits of predictive analytics is outlined in a sperate article from DPS.
Ref: 1. Article in International Journal of Global Energy Issues 35(6):518 - 533 · January 2012. Authors - Roukia Ouddai, Hassane Chabane, Assia Boughaba and Mohamed Frah.