Driving reliability and improving maintenance outcomes with machine learning

In 2017, McKinsey conducted a study on productivity gains driven by technology transformations, such as the steam engine, early robotic technology and advances in information technology. McKinsey sees manufacturing on the brink of the next generation of industrial automation revolution with unprecedented annual productivity growth of between 0.8 – 1.4% in the decades ahead. Advances in robotics, artificial intelligence and machine learning will match or outperform humans in a range of work activities involving fast, precise, repetitive action and cognitive capabilities. To remain competitive, complex industries need to deploy industrial automation more than ever, as intense global competition drives process industries to increase efficiency through reduced operating costs, increased production, higher quality and lower inventories. The highest priority should be to eliminate production losses caused by unplanned downtime and address a $20 billion a year problem for the process industries. As such, increased asset utilisation will bring the single biggest financial improvement in production operations.  

Current Maintenance Practices   

For the last approximately 50 years, maintenance practices have evolved in terms of equipment reliability and availability. Maintenance strategy has progressed through run-to-failure; calendar-based; usage-based; condition-based and reliability-centered maintenance. Outcomes have improved but the equipment continues to fail. Why? Despite each successive technique becoming more complex, they are not addressing the main issue. Industry analysts, such as ARC Advisory Group, have pointed out that more than 80% of all equipment failures are caused by operating equipment outside its stipulated design and safety limits. Current practices that only focus on 20% of the issues involved cannot detect problems early enough, lack insight into the reasons behind the seemingly random failures and cannot detect seemingly “random” equipment failures. Manufacturers need solutions that arrive at the confluence of both maintenance and operations activities to address 100% of failures.  

Significantly, these solutions need to offer failure prevention using data-driven truths rather than guesstimates. The combination of mechanical and process induced breakdowns costs up to 10 per cent of a worldwide $1.4 trillion manufacturing market per a 2012 report from The McKinsey Global Institute. While companies have spent millions trying to address this issue and ultimately avoid unplanned downtime, until now, they have only been able to address wear and age-based failures. This is where using machine learning software to cast a “wider net” around machines can capture process induced failures.    

To avoid unplanned downtime, companies must identify and respond effectively to early indicators of impending failures. Traditional maintenance practices do not predict failures caused by process excursions. That would require a unique technology approach combining machines and processes; particularly for asset-intensive industries such as manufacturing and transportation. With the right technology in place, organisations can sense the patterns of looming degradation, with sufficient warning to prevent failures and change outcomes.  

Predicting Downtime with Machine Learning Software 

Advanced machine learning software has already demonstrated incredible successes in the early identification of equipment failure. Such software is near-autonomous and learns behavioural patterns from the streams of digital data that are produced by sensors on and around machines and processes. Automatically, and requiring minimal resources, this advanced technology constantly learns and adapts to new signal patterns when operating conditions change. Failure signatures learned on one machine “inoculate” that machine so that the same condition will not recur. Additionally, the learned signatures are transferred to similar machines to prevent them being affected by the same degrading conditions.    

For example, a North American energy company was losing up to a million dollars in repairs and lost revenue from repeat breakdowns of electric submersible pumps. The advanced machine learning application learned the behaviour of 18 pumps. The software detected an early casing leak on one pump that caused an environmental incident. Applying the failure signature to the rest of the pumps provided an early warning, allowing early action to avoid a repeat incident, thus preventing a major problem. 

In another case, a leading railway freight firm operating across 23 states in the US used machine learning to address perennial locomotive engine failures costing millions in repairs, fines and lost revenue. The machine learning application operates in-line, in real-time and was deployed on a very large fleet of locomotives examining lube oil data for extremely early indicators of engine failure. The application even detected a degradation signature while the engine passed a low-pressure test. Diverting the locomotive for immediate service “saved the company millions of dollars in costly downtime and fines.” 

The Next Generation Asset Performance Management (APM) 2.0 

ARC Advisory Group has further asserted that APM 2.0 incorporates new analytics and data from control systems and asset management applications, providing new opportunities to optimise availability and operational performance. APM 2.0 strategies include sharing information from other systems, such as manufacturing execution systems, to deliver a comprehensive view and analysis of production processes and asset performance. Leading edge analytics view data from all systems to assess patterns of normal and abnormal behaviour, which helps to predict future conditions and their inherent causes. Such activities promote deeper collaboration between operations and maintenance staff via intertwined decision making, driven by shared goals. The combination of data supports shared and improved understanding of risk, which provides the ability to balance operational constraints and efficiency opportunities to improve return on assets.    

The Time to Implement Machine Learning Software is Now  

Companies can no longer rely solely on traditional maintenance practices but must also incorporate operational behaviours in deploying data-driven solutions. Today’s imperative means extracting additional value from existing assets and implementing an advanced machine learning programme to deliver fast improvements. With the right software solutions, predictive technologies will detect the conditions that limit asset effectiveness, while providing prescriptive guidance that assures firms remain profitable and improve margins.    

Mike Brooks, Senior Business Consultant, AspenTech and Former Mtell President & Chief Operating Officer 

Image Credit: Sergey Nivens / Shutterstock