It’s time to move from ‘fail & fix’ to ‘predict & prevent’

Predictive maintenance approaches have reduced maintenance cost by over 25 per cent and increased production by over 20 per cent (O&M Best Practices Guide 2010).

In today’s volatile and complex global economic situation, companies refraining from capital investment are forced to get more out of their physical asset base. With increasing pressure to reduce operational costs, industries find it difficult to justify the fixed cost of preventive maintenance and machine spares used in Total Productive Maintenance (TPM).

Production managers do not have many tools to ensure reliability and smooth running of facilities. In the absence of reliable data or methodology, the behavior of machines over a period cannot be accurately predicted. Investments, thus, are based on experience and intuition. This calls for new ways to manage physical assets and maintenance practices. It’s time to move away from the traditional ‘fail & fix’ or the ‘fix whether it’s broken or not’ approaches. It is time to adopt a data-centric analytical approach based on a ‘predict-and-prevent’ paradigm. This will bring about predictability in operations and achieve near zero downtime that can reduce the cost.

Predictive maintenance approaches have recently been implemented in process industries like oil & gas, power plants and wind turbines. As per the Operations &Maintenance Best Practices Guide 2010, these approaches have reduced maintenance cost by 25-30 per cent and increased production by 20-25 per cent. Predictive maintenance program can provide better ROI (approximately 10 times higher) and a saving of 8-12 per cent over the preventive maintenance program.

The predictive maintenance approach employs collection of machine data using sensors and industrial devices. The data is analyzed via complex algorithms and patterns are established that help predict the probable behavior of the machine in terms of its functionality, quality and maintenance. Manufacturing systems are already generating a huge amount of production data, through computer integrated manufacturing systems. The decreasing cost of sensors, ubiquitous access to the internet and computer integrated manufacturing allows the exchange of data between machine, work-piece, system and people.

Powerful data-driven technologies like Big Data Analytics, through cheaper processing power, can help companies generate new insights and solve complex problems. The digital IQ Survey of PricewaterhouseCoopers (PwC) showed that 54 per cent of top performing businesses are adding sensors to people, places, processes and products to gather and analyze information that helps them make better decisions and increase transparency.

Though predictive maintenance has brought in considerable improvement in maintaining asset efficiency and ensuring reliable production in the process industry, it is yet to be implemented in discrete manufacturing industries to any significant extent. As mentioned by TCS in a white paper on big data in machine learning analytics, some of the challenges in implementing this approach in discrete environment are:

a) Lack of expertise – few data scientists have experience in the manufacturing industry
b) Lack of industrial readiness
c) Heterogeneous asset condition, leading to a lack of availability of comparable, consistent data
d) Openness to long term investments due to cost pressure in manufacturing environment

Successful implementation of predictive maintenance requires commitment from the top management to enable changes in the process. Technology and analytics capability coupled with strong domain knowledge can bring about a tremendous change in operations. For instance, an automobile component manufacturer improved its overall equipment effectiveness (OEE) by 15 per cent and its energy cost by 10 per cent, using analytics approaches in maintenance.

At CGN, we have been working on connecting the threads and applying them to tap into the advantages of predictive maintenance in lowering TPM costs.